add image_classification

model_zoo/official/lite/.gitignore

@@ -0,0 +1,81 @@
+# MindSpore
+build/
+mindspore/lib
+output
+*.ir
+mindspore/ccsrc/schema/inner/*
+
+# Cmake files
+CMakeFiles/
+cmake_install.cmake
+CMakeCache.txt
+Makefile
+cmake-build-debug
+
+# Dynamic libraries
+*.so
+*.so.*
+*.dylib
+
+# Static libraries
+*.la
+*.lai
+*.a
+*.lib
+
+# Protocol buffers
+*_pb2.py
+*.pb.h
+*.pb.cc
+
+# Object files
+*.o
+
+# Editor
+.vscode
+.idea/
+
+# Cquery
+.cquery_cached_index/
+compile_commands.json
+
+# Ctags and cscope
+tags
+TAGS
+CTAGS
+GTAGS
+GRTAGS
+GSYMS
+GPATH
+cscope.*
+
+# Python files
+*__pycache__*
+.pytest_cache
+
+# Mac files
+*.DS_Store
+
+# Test results
+test_temp_summary_event_file/
+*.dot
+*.dat
+*.svg
+*.perf
+*.info
+*.ckpt
+*.shp
+*.pkl
+.clangd
+mindspore/version.py
+mindspore/default_config.py
+mindspore/.commit_id
+onnx.proto
+mindspore/ccsrc/onnx.proto
+
+# Android
+local.properties
+.gradle
+sdk/build
+sdk/.cxx
+app/.cxx

model_zoo/official/lite/README.md

@@ -0,0 +1,275 @@
+## MindSpore Lite 端侧图像分类demo（Android）
+  
+本示例程序演示了如何在端侧利用MindSpore Lite C++ API（Android JNI）以及MindSpore Lite 图像分类模型完成端侧推理，实现对设备摄像头捕获的内容进行分类，并在App图像预览界面中显示出最可能的分类结果。
+
+
+### 运行依赖
+
+- Android Studio >= 3.2 (推荐4.0以上版本)
+- NDK 21.3
+- CMake 3.10
+- Android SDK >= 26
+- OpenCV >= 4.0.0
+
+### 构建与运行
+
+1. 在Android Studio中加载本示例源码，并安装相应的SDK（指定SDK版本后，由Android Studio自动安装）。 
+
+    ![start_home](images/home.png)
+
+    启动Android Studio后，点击`File->Settings->System Settings->Android SDK`，勾选相应的SDK。如下图所示，勾选后，点击`OK`，Android Studio即可自动安装SDK。
+
+    ![start_sdk](images/sdk_management.png)
+
+    （可选）若安装时出现NDK版本问题，可手动下载相应的[NDK版本](https://developer.android.com/ndk/downloads?hl=zh-cn)（本示例代码使用的NDK版本为21.3），并在`Project Structure`的`Android NDK location`设置中指定SDK的位置。
+
+    ![project_structure](images/project_structure.png)
+
+2. 连接Android设备，运行图像分类应用程序。
+
+    通过USB连接Android设备调试，点击`Run 'app'`即可在您的设备上运行本示例项目。
+
+    * 注：编译过程中Android Studio会自动下载MindSpore Lite、OpenCV、模型文件等相关依赖项，编译过程需做耐心等待。
+
+    ![run_app](images/run_app.PNG)
+
+    Android Studio连接设备调试操作，可参考<https://developer.android.com/studio/run/device?hl=zh-cn>。
+
+3. 在Android设备上，点击“继续安装”，安装完即可查看到设备摄像头捕获的内容和推理结果。
+
+    ![install](images/install.jpg)
+
+    如下图所示，识别出的概率最高的物体是植物。
+
+    ![result](images/app_result.jpg)
+
+
+## 示例程序详细说明  
+
+本端侧图像分类Android示例程序分为JAVA层和JNI层，其中，JAVA层主要通过Android Camera 2 API实现摄像头获取图像帧，以及相应的图像处理等功能；JNI层在[Runtime](https://www.mindspore.cn/tutorial/zh-CN/master/use/lite_runtime.html)中完成模型推理的过程。
+
+> 此处详细说明示例程序的JNI层实现，JAVA层运用Android Camera 2 API实现开启设备摄像头以及图像帧处理等功能，需读者具备一定的Android开发基础知识。
+
+### 示例程序结构
+
+```
+app
+|
+├── libs # 存放demo jni层依赖的库文件
+│   └── arm64-v8a
+│       ├── libopencv_java4.so  # opencv
+│       ├── libmlkit-label-MS.so  # ndk编译生成的库文件
+│       └── libmindspore-lite.so # mindspore lite
+|
+├── src/main
+│   ├── assets # 资源文件
+|   |   └── mobilenetv2.ms # 存放模型文件
+│   |
+│   ├── cpp # 模型加载和预测主要逻辑封装类
+|   |   ├── include # 存放MindSpore调用相关的头文件
+|   |   |   └── ...
+│   |   |
+|   |   ├── MindSporeNetnative.cpp # MindSpore调用相关的JNI方法
+│   |   └── MindSporeNetnative.h # 头文件
+│   |
+│   ├── java # java层应用代码
+│   │   └── com.huawei.himindsporedemo 
+│   │       ├── gallery.classify # 图像处理及MindSpore JNI调用相关实现
+│   │       │   └── ...
+│   │       └── obejctdetect # 开启摄像头及绘制相关实现
+│   │           └── ...
+│   │   
+│   ├── res # 存放Android相关的资源文件
+│   └── AndroidManifest.xml # Android配置文件
+│
+├── CMakeList.txt # cmake编译入口文件
+│
+├── build.gradle # 其他Android配置文件
+├── download.gradle # APP构建时由gradle自动从HuaWei Server下载依赖的库文件及模型文件
+└── ...
+```
+
+### 配置MindSpore Lite依赖项
+
+Android JNI层调用MindSpore C++ API时，需要相关库文件支持。可通过MindSpore Lite[源码编译](https://www.mindspore.cn/lite/docs/zh-CN/master/deploy.html)生成`libmindspore-lite.so`库文件，或直接下载MindSpore Lite提供的已编译完成的AMR64、ARM32、x86等[软件包](#TODO)。
+
+在Android Studio中将编译完成的`libmindspore-lite.so`库文件（可包含多个兼容架构），分别放置在APP工程的`app/libs/arm64-v8a`（ARM64）或`app/libs/armeabi-v7a`（ARM32）目录下，并在应用的`build.gradle`文件中配置CMake编译支持，以及`arm64-v8a`和`armeabi-v7a`的编译支持。
+
+本示例中，build过程由download.gradle文件自动从华为服务器下载libmindspore-lite.so以及OpenCV的libopencv_java4.so库文件，并放置在`app/libs/arm64-v8a`目录下。
+
+* 注：若自动下载失败，请手动下载相关库文件并将其放在对应位置：
+* libmindspore-lite.so [下载链接](https://download.mindspore.cn/model_zoo/official/lite/lib/mindspore%20version%200.7/libmindspore-lite.so)
+* libopencv_java4.so [下载链接](https://download.mindspore.cn/model_zoo/official/lite/lib/opencv%204.4.0/libopencv_java4.so)
+
+```
+android{
+    defaultConfig{
+        externalNativeBuild{
+            cmake{
+                arguments "-DANDROID_STL=c++_shared"
+            }
+        }
+
+        ndk{ 
+            abiFilters 'arm64-v8a'
+        }
+    }
+}
+```
+
+在`app/CMakeLists.txt`文件中建立`.so`库文件链接，如下所示。
+
+```
+# Set MindSpore Lite Dependencies.
+include_directories(${CMAKE_SOURCE_DIR}/src/main/cpp/include/MindSpore)
+add_library(mindspore-lite SHARED IMPORTED )
+set_target_properties(mindspore-lite PROPERTIES
+    IMPORTED_LOCATION "${CMAKE_SOURCE_DIR}/libs/libmindspore-lite.so")
+
+# Set OpenCV Dependecies.
+include_directories(${CMAKE_SOURCE_DIR}/opencv/sdk/native/jni/include)
+add_library(lib-opencv SHARED IMPORTED )
+set_target_properties(lib-opencv PROPERTIES
+    IMPORTED_LOCATION "${CMAKE_SOURCE_DIR}/libs/libopencv_java4.so")
+
+# Link target library.       
+target_link_libraries(
+    ...
+    mindspore-lite
+    lib-opencv
+    ...
+)
+```
+
+### 下载及部署模型文件
+
+从MindSpore Model Hub中下载模型文件，本示例程序中使用的终端图像分类模型文件为`mobilenetv2.ms`，同样通过download.gradle脚本在APP构建时自动下载，并放置在`app/src/main/assets`工程目录下。
+
+* 注：若下载失败请手动下载模型文件，mobilenetv2.ms [下载链接](https://download.mindspore.cn/model_zoo/official/lite/mobilenetv2_openimage_lite/mobilenetv2.ms)。
+
+
+### 编写端侧推理代码
+
+在JNI层调用MindSpore Lite C++ API实现端测推理。
+
+推理代码流程如下，完整代码请参见`src/cpp/MindSporeNetnative.cpp`。 
+
+1. 加载MindSpore Lite模型文件，构建上下文、会话以及用于推理的计算图。  
+
+    - 加载模型文件：创建并配置用于模型推理的上下文
+        ```cpp
+        // Buffer is the model data passed in by the Java layer
+        jlong bufferLen = env->GetDirectBufferCapacity(buffer);
+        char *modelBuffer = CreateLocalModelBuffer(env, buffer);  
+        ```
+        
+    - 创建会话
+        ```cpp
+        void **labelEnv = new void *;
+        MSNetWork *labelNet = new MSNetWork;
+        *labelEnv = labelNet;
+        
+        // Create context.
+        lite::Context *context = new lite::Context;
+        context->thread_num_ = numThread;  //Specify the number of threads to run inference
+        
+        // Create the mindspore session.
+        labelNet->CreateSessionMS(modelBuffer, bufferLen, context);
+        delete(context);
+        
+        ```
+        
+    - 加载模型文件并构建用于推理的计算图
+        ```cpp
+        void MSNetWork::CreateSessionMS(char* modelBuffer, size_t bufferLen, mindspore::lite::Context* ctx)
+        {
+            CreateSession(modelBuffer, bufferLen, ctx);  
+            session = mindspore::session::LiteSession::CreateSession(ctx);
+            auto model = mindspore::lite::Model::Import(modelBuffer, bufferLen);
+            int ret = session->CompileGraph(model); // Compile Graph 
+        }
+        ```
+    
+2. 将输入图片转换为传入MindSpore模型的Tensor格式。 
+
+    将待检测图片数据转换为输入MindSpore模型的Tensor。
+
+    ```cpp
+    // Convert the Bitmap image passed in from the JAVA layer to Mat for OpenCV processing
+    BitmapToMat(env, srcBitmap, matImageSrc);
+   // Processing such as zooming the picture size.
+    matImgPreprocessed = PreProcessImageData(matImageSrc);  
+
+    ImgDims inputDims; 
+    inputDims.channel = matImgPreprocessed.channels();
+    inputDims.width = matImgPreprocessed.cols;
+    inputDims.height = matImgPreprocessed.rows;
+    float *dataHWC = new float[inputDims.channel * inputDims.width * inputDims.height]
+
+    // Copy the image data to be detected to the dataHWC array.
+    // The dataHWC[image_size] array here is the intermediate variable of the input MindSpore model tensor.
+    float *ptrTmp = reinterpret_cast<float *>(matImgPreprocessed.data);
+    for(int i = 0; i < inputDims.channel * inputDims.width * inputDims.height; i++){
+       dataHWC[i] = ptrTmp[i];
+    }
+
+    // Assign dataHWC[image_size] to the input tensor variable.
+    auto msInputs = mSession->GetInputs();
+    auto inTensor = msInputs.front();
+    memcpy(inTensor->MutableData(), dataHWC,
+        inputDims.channel * inputDims.width * inputDims.height * sizeof(float));
+    delete[] (dataHWC);
+    ```
+    
+3. 对输入Tensor按照模型进行推理，获取输出Tensor，并进行后处理。    
+
+   - 图执行，端测推理。
+
+        ```cpp
+        // After the model and image tensor data is loaded, run inference.
+        auto status = mSession->RunGraph();
+        ```
+
+   - 获取输出数据。
+        ```cpp
+        // Get the mindspore inference results.
+        auto msOutputs = mSession->GetOutputMapByNode();
+        std::string retStr = ProcessRunnetResult(msOutputs);
+        ```
+        
+   - 输出数据的后续处理。
+        ```cpp
+        std::string ProcessRunnetResult(
+                std::unordered_map<std::string, std::vector<mindspore::tensor::MSTensor *>> msOutputs){
+        
+            // Get the branch of the model output.
+            // Use iterators to get map elements.
+            std::unordered_map<std::string, std::vector<mindspore::tensor::MSTensor *>>::iterator iter;
+            iter = msOutputs.begin();
+
+            // The mobilenetv2.ms model output just one branch.
+            auto outputString = iter->first;
+            auto outputTensor = iter->second;
+
+            float *temp_scores = static_cast<float * >(branch1_tensor[0]->MutableData());
+        
+            float scores[RET_CATEGORY_SUM];
+            for (int i = 0; i < RET_CATEGORY_SUM; ++i) {
+                if (temp_scores[i] > 0.5){
+                    MS_PRINT("MindSpore scores[%d] : [%f]",  i, temp_scores[i]);
+                }
+                scores[i] = temp_scores[i];
+            }
+   
+            // Converted to text information that needs to be displayed in the APP. 
+            std::string categoryScore = "";
+                for (int i = 0; i < RET_CATEGORY_SUM; ++i) {
+                    categoryScore += g_labels_name_map[i];
+                    categoryScore += ":";
+                    std::string score_str = std::to_string(scores[i]);
+                    categoryScore += score_str;
+                    categoryScore += ";";
+                }
+                return categoryScore;
+        }      
+        ```

model_zoo/official/lite/app/.gitignore

@@ -0,0 +1 @@
+/build

model_zoo/official/lite/app/CMakeLists.txt

@@ -0,0 +1,94 @@
+# For more information about using CMake with Android Studio, read the
+# documentation: https://d.android.com/studio/projects/add-native-code.html
+
+# Sets the minimum version of CMake required to build the native library.
+
+cmake_minimum_required(VERSION 3.4.1)
+
+set(CMAKE_VERBOSE_MAKEFILE on)
+set(libs  ${CMAKE_SOURCE_DIR}/libs)
+
+
+set(CMAKE_LIBRARY_OUTPUT_DIRECTORY  ${CMAKE_SOURCE_DIR}/libs/${ANDROID_ABI})
+
+
+# ============== Set MindSpore Dependencies. =============
+include_directories(${CMAKE_SOURCE_DIR}/src/main/cpp/include)
+include_directories(${CMAKE_SOURCE_DIR}/src/main/cpp/include/MindSpore)
+include_directories(${CMAKE_SOURCE_DIR}/src/main/cpp/include/MindSpore/flatbuffers)
+include_directories(${CMAKE_SOURCE_DIR}/src/main/cpp/include/MindSpore/ir/dtype)
+include_directories(${CMAKE_SOURCE_DIR}/src/main/cpp/include/MindSpore/schema)
+
+add_library(mindspore-lite SHARED IMPORTED )
+
+set_target_properties(mindspore-lite PROPERTIES IMPORTED_LOCATION
+        ${CMAKE_SOURCE_DIR}/libs/${ANDROID_ABI}/libmindspore-lite.so)
+# --------------- MindSpore Lite set End. --------------------
+
+
+
+# =============== Set OpenCV Dependencies ===================
+
+include_directories(${CMAKE_SOURCE_DIR}/opencv/sdk/native/jni/include/)
+
+add_library(lib-opencv SHARED IMPORTED )
+
+set_target_properties(lib-opencv PROPERTIES IMPORTED_LOCATION
+        ${CMAKE_SOURCE_DIR}/libs/${ANDROID_ABI}/libopencv_java4.so)
+
+# --------------- OpenCV set End. ---------------------------
+
+
+# Creates and names a library, sets it as either STATIC
+# or SHARED, and provides the relative paths to its source code.
+# You can define multiple libraries, and CMake builds them for you.
+# Gradle automatically packages shared libraries with your APK.
+
+file(GLOB_RECURSE cpp_src "src/main/cpp/*.cpp"  "src/main/cpp/*.h")
+
+add_library( # Sets the name of the library.
+        mlkit-label-MS
+
+        # Sets the library as a shared library.
+        SHARED
+
+        # Provides a relative path to your source file(s).
+        ${cpp_src})
+
+
+# Searches for a specified prebuilt library and stores the path as a
+# variable. Because CMake includes system libraries in the search path by
+# default, you only need to specify the name of the public NDK library
+# you want to add. CMake verifies that the library exists before
+# completing its build.
+
+find_library( # Sets the name of the path variable.
+        log-lib
+
+        # Specifies the name of the NDK library that
+        # you want CMake to locate.
+        log )
+
+
+find_library( jnigraphics-lib jnig·raphics )
+
+# Specifies libraries CMake should link to your target library. You
+# can link multiple libraries, such as libraries you define in this
+# build script, prebuilt third-party libraries, or system libraries.
+add_definitions(-DMNN_USE_LOGCAT)
+target_link_libraries( # Specifies the target library.
+        mlkit-label-MS
+
+        # --- opencv ---
+        lib-opencv
+
+        # --- mindspore ---
+        mindspore-lite
+
+        # --- other dependencies.---
+        -ljnigraphics
+        android
+
+        # Links the target library to the log library
+        ${log-lib}
+        )

model_zoo/official/lite/app/build.gradle

@@ -0,0 +1,84 @@
+apply plugin: 'com.android.application'
+
+android {
+    compileSdkVersion 30
+    buildToolsVersion "30.0.1"
+
+    defaultConfig {
+        applicationId "com.huawei.himindsporedemo"
+        minSdkVersion 21
+        targetSdkVersion 30
+        versionCode 1
+        versionName "1.0"
+
+        testInstrumentationRunner "androidx.test.runner.AndroidJUnitRunner"
+        externalNativeBuild {
+            cmake {
+                arguments "-DANDROID_STL=c++_shared"
+                cppFlags ""
+            }
+        }
+        ndk {
+            abiFilters 'arm64-v8a'
+        }
+    }
+    aaptOptions {
+        noCompress '.so', 'ms'
+    }
+
+    buildTypes {
+        release {
+            minifyEnabled false
+            proguardFiles getDefaultProguardFile('proguard-android-optimize.txt'), 'proguard-rules.pro'
+        }
+
+        customDebugType {
+            debuggable true
+        }
+    }
+    externalNativeBuild {
+        cmake {
+            path file('CMakeLists.txt')
+        }
+    }
+    ndkVersion '21.3.6528147'
+
+    sourceSets{
+        main {
+            jniLibs.srcDirs = ['libs']
+        }
+    }
+    packagingOptions{
+        pickFirst 'lib/arm64-v8a/libopencv_java4.so'
+        pickFirst 'lib/arm64-v8a/libmindspore-lite.so'
+        pickFirst 'lib/arm64-v8a/libmlkit-label-MS.so'
+    }
+
+}
+
+
+
+// Before gradle build.
+// To download some necessary libraries.
+apply from:'download.gradle'
+/*if (!file("libs/arm64-v8a/libmindspore-lite.so").exists() ||
+        !file("libs/arm64-v8a/libopencv_java4.so").exists()){
+    apply from:'download.gradle'
+}*/
+
+
+
+dependencies {
+    implementation fileTree(dir: "libs", include: ["*.jar"])
+    implementation 'androidx.appcompat:appcompat:1.1.0'
+    implementation 'androidx.constraintlayout:constraintlayout:1.1.3'
+//    implementation project(path: ':sdk')
+
+    testImplementation 'junit:junit:4.12'
+    androidTestImplementation 'androidx.test.ext:junit:1.1.1'
+    androidTestImplementation 'androidx.test.espresso:espresso-core:3.2.0'
+
+    implementation 'com.google.android.material:material:1.0.0'
+    androidTestImplementation 'com.android.support.test:rules:1.0.2'
+    androidTestImplementation 'com.google.truth:truth:1.0.1'
+}

model_zoo/official/lite/app/download.gradle

@@ -0,0 +1,73 @@
+/**
+ * To download necessary library from HuaWei server.
+ * Including mindspore-lite .so file, opencv .so file and model file.
+ * The libraries can be downloaded manually.
+ */
+
+
+def targetModelFile = "src/main/assets/model/mobilenetv2.ms"
+def openCVLibrary_arm64 = "libs/arm64-v8a/libopencv_java4.so"
+def mindSporeLibrary_arm64 = "libs/arm64-v8a/libmindspore-lite.so"
+
+def modelDownloadUrl = "https://download.mindspore.cn/model_zoo/official/lite/mobilenetv2_openimage_lite/mobilenetv2.ms"
+def opencvDownloadUrl = "https://download.mindspore.cn/model_zoo/official/lite/lib/opencv%204.4.0/libopencv_java4.so"
+def mindsporeLiteDownloadUrl = "https://download.mindspore.cn/model_zoo/official/lite/lib/mindspore%20version%200.7/libmindspore-lite.so"
+
+
+task downloadModelFile(type: DownloadUrlTask) {
+    doFirst {
+        println "Downloading ${modelDownloadUrl}"
+    }
+    sourceUrl = "${modelDownloadUrl}"
+    target = file("${targetModelFile}")
+}
+
+
+task downloadOpenCVLibrary(type: DownloadUrlTask) {
+    doFirst {
+        println "Downloading ${opencvDownloadUrl}"
+    }
+    sourceUrl = "${opencvDownloadUrl}"
+    target = file("${openCVLibrary_arm64}")
+}
+
+task downloadMindSporeLibrary(type: DownloadUrlTask) {
+    doFirst {
+        println "Downloading ${mindsporeLiteDownloadUrl}"
+    }
+    sourceUrl = "${mindsporeLiteDownloadUrl}"
+    target = file("${mindSporeLibrary_arm64}")
+}
+
+/*
+ * Using preBuild to download mindspore library, opencv library and model file.
+ * Run before gradle build.
+ */
+if (file("libs/arm64-v8a/libmindspore-lite.so").exists()){
+    downloadMindSporeLibrary.enabled = false
+}
+
+if (file("libs/arm64-v8a/libopencv_java4.so.so").exists()){
+    downloadOpenCVLibrary.enabled = false
+}
+if (file("src/main/assets/model/mobilenetv2.ms").exists()){
+    downloadModelFile.enabled = false
+}
+
+preBuild.dependsOn downloadMindSporeLibrary
+preBuild.dependsOn downloadOpenCVLibrary
+preBuild.dependsOn downloadModelFile
+
+
+class DownloadUrlTask extends DefaultTask {
+    @Input
+    String sourceUrl
+
+    @OutputFile
+    File target
+
+    @TaskAction
+    void download() {
+        ant.get(src: sourceUrl, dest: target)
+    }
+}

model_zoo/official/lite/app/proguard-rules.pro

@@ -0,0 +1,21 @@
+# Add project specific ProGuard rules here.
+# You can control the set of applied configuration files using the
+# proguardFiles setting in build.gradle.
+#
+# For more details, see
+#   http://developer.android.com/guide/developing/tools/proguard.html
+
+# If your project uses WebView with JS, uncomment the following
+# and specify the fully qualified class name to the JavaScript interface
+# class:
+#-keepclassmembers class fqcn.of.javascript.interface.for.webview {
+#   public *;
+#}
+
+# Uncomment this to preserve the line number information for
+# debugging stack traces.
+#-keepattributes SourceFile,LineNumberTable
+
+# If you keep the line number information, uncomment this to
+# hide the original source file name.
+#-renamesourcefileattribute SourceFile

model_zoo/official/lite/app/src/androidTest/java/com/huawei/himindsporedemo/ExampleInstrumentedTest.java

@@ -0,0 +1,26 @@
+package com.huawei.himindsporedemo;
+
+import android.content.Context;
+
+import androidx.test.platform.app.InstrumentationRegistry;
+import androidx.test.ext.junit.runners.AndroidJUnit4;
+
+import org.junit.Test;
+import org.junit.runner.RunWith;
+
+import static org.junit.Assert.*;
+
+/**
+ * Instrumented test, which will execute on an Android device.
+ *
+ * @see <a href="http://d.android.com/tools/testing">Testing documentation</a>
+ */
+@RunWith(AndroidJUnit4.class)
+public class ExampleInstrumentedTest {
+    @Test
+    public void useAppContext() {
+        // Context of the app under test.
+        Context appContext = InstrumentationRegistry.getInstrumentation().getTargetContext();
+        assertEquals("com.huawei.himindsporedemo", appContext.getPackageName());
+    }
+}

model_zoo/official/lite/app/src/main/AndroidManifest.xml

@@ -0,0 +1,32 @@
+<?xml version="1.0" encoding="utf-8"?>
+<manifest xmlns:android="http://schemas.android.com/apk/res/android"
+    package="com.huawei.himindsporedemo"
+    android:versionCode="1"
+    android:versionName="1.0">
+
+    <uses-permission android:name="android.permission.CAMERA" />
+    <uses-permission android:name="android.permission.WRITE_INTERNAL_STORAGE" />
+    <uses-permission android:name="android.permission.READ_INTERNAL_STORAGE" />
+    <uses-permission android:name="android.permission.READ_EXTERNAL_STORAGE" />
+    <uses-permission android:name="android.permission.WRITE_EXTERNAL_STORAGE" />
+    <uses-permission android:name="android.permission.MOUNT_UNMOUNT_FILESYSTEM" />
+    <uses-permission android:name="android.permission.READ_PHONE_STATE" />
+
+    <application
+        android:allowBackup="true"
+        android:icon="@mipmap/ic_launcher"
+        android:label="@string/app_name"
+        android:roundIcon="@mipmap/ic_launcher_round"
+        android:supportsRtl="true"
+        android:theme="@style/AppTheme">
+        <activity
+            android:name=".widget.CameraActivity"
+            android:screenOrientation="portrait">
+            <intent-filter>
+                <action android:name="android.intent.action.MAIN" />
+                <category android:name="android.intent.category.LAUNCHER" />
+            </intent-filter>
+        </activity>
+    </application>
+
+</manifest>

model_zoo/official/lite/app/src/main/cpp/MindSporeNetnative.cpp

@@ -0,0 +1,296 @@
+/*
+ * Copyright (c) Huawei Technologies Co., Ltd. 2018-2019. All rights reserved.
+ */
+#include <android/bitmap.h>
+#include <android/asset_manager_jni.h>
+#include <android/log.h>
+#include <jni.h>
+#include <cstring>
+#include <set>
+
+#include <MindSpore/errorcode.h>
+#include <MindSpore/ms_tensor.h>
+#include "MindSporeNetnative.h"
+#include "opencv2/core.hpp"
+#include "opencv2/imgproc.hpp"
+#include "MindSpore/MSNetWork.h"
+#include "HMS/HMS_label_thres.h"
+
+using namespace cv;
+using namespace mindspore;
+using namespace mindspore::tensor;
+
+#define MS_PRINT(format, ...) __android_log_print(ANDROID_LOG_INFO, "MSJNI", format, ##__VA_ARGS__)
+
+
+void BitmapToMat2(JNIEnv *env, jobject &bitmap, Mat &mat, jboolean needUnPremultiplyAlpha) {
+    AndroidBitmapInfo info;
+    void *pixels = nullptr;
+    Mat &dst = mat;
+    CV_Assert(AndroidBitmap_getInfo(env, bitmap, &info) >= 0);
+    CV_Assert(info.format == ANDROID_BITMAP_FORMAT_RGBA_8888 ||
+              info.format == ANDROID_BITMAP_FORMAT_RGB_565);
+    CV_Assert(AndroidBitmap_lockPixels(env, bitmap, &pixels) >= 0);
+    CV_Assert(pixels);
+
+    dst.create(info.height, info.width, CV_8UC4);
+    if (info.format == ANDROID_BITMAP_FORMAT_RGBA_8888) {
+        Mat tmp(info.height, info.width, CV_8UC4, pixels);
+        if (needUnPremultiplyAlpha) {
+            cvtColor(tmp, dst, COLOR_RGBA2BGR);
+        } else {
+
+            tmp.copyTo(dst);
+        }
+    } else {
+        Mat tmp(info.height, info.width, CV_8UC4, pixels);
+        cvtColor(tmp, dst, COLOR_BGR5652RGBA);
+    }
+    AndroidBitmap_unlockPixels(env, bitmap);
+    return;
+}
+
+void BitmapToMat(JNIEnv *env, jobject &bitmap, Mat &mat) {
+    BitmapToMat2(env, bitmap, mat, true);
+}
+
+/**
+ * Processing image with resize and normalize.
+ */
+cv::Mat PreProcessImageData(cv::Mat input) {
+    cv::Mat imgFloatTmp, imgResized256, imgResized224;
+    int resizeWidth = 256;
+    int resizeHeight = 256;
+    float normalizMin = 1.0;
+    float normalizMax = 255.0;
+
+    cv::resize(input, imgFloatTmp, cv::Size(resizeWidth, resizeHeight));
+
+
+    imgFloatTmp.convertTo(imgResized256, CV_32FC3, normalizMin / normalizMax);
+
+    int offsetX = 16;
+    int offsetY = 16;
+    int cropWidth = 224;
+    int cropHeight = 224;
+
+    // Standardization processing.
+    float meanR = 0.485;
+    float meanG = 0.456;
+    float meanB = 0.406;
+    float varR = 0.229;
+    float varG = 0.224;
+    float varB = 0.225;
+
+    cv::Rect roi;
+    roi.x = offsetX;
+    roi.y = offsetY;
+    roi.width = cropWidth;
+    roi.height = cropHeight;
+
+    // The final image size of the incoming model is 224*224.
+    imgResized256(roi).copyTo(imgResized224);
+
+    Scalar mean = Scalar(meanR, meanG, meanB);
+    Scalar var = Scalar(varR, varG, varB);
+    cv::Mat imgResized1;
+    cv::Mat imgResized2;
+    Mat imgMean(imgResized224.size(), CV_32FC3,
+                mean);  // imgMean Each pixel channel is (0.485, 0.456, 0.406)
+    Mat imgVar(imgResized224.size(), CV_32FC3,
+               var); // imgVar Each pixel channel is (0.229, 0.224, 0.225)
+    imgResized1 = imgResized224 - imgMean;
+    imgResized2 = imgResized1 / imgVar;
+    return imgResized2;
+}
+
+char *CreateLocalModelBuffer(JNIEnv *env, jobject modelBuffer) {
+    jbyte *modelAddr = static_cast<jbyte *>(env->GetDirectBufferAddress(modelBuffer));
+    int modelLen = static_cast<int>(env->GetDirectBufferCapacity(modelBuffer));
+    char *buffer(new char[modelLen]);
+    memcpy(buffer, modelAddr, modelLen);
+    return buffer;
+}
+
+/**
+ * To process the result of mindspore inference.
+ * @param msOutputs
+ * @return
+ */
+std::string ProcessRunnetResult(
+        std::unordered_map<std::string, std::vector<mindspore::tensor::MSTensor *>> msOutputs) {
+
+    // Get the branch of the model output.
+    // Use iterators to get map elements.
+    std::unordered_map<std::string, std::vector<mindspore::tensor::MSTensor *>>::iterator iter;
+    iter = msOutputs.begin();
+
+    // The mobilenetv2.ms model output just one branch.
+    auto outputString = iter->first;
+    auto outputTensor = iter->second;
+
+    int tensorNum = outputTensor[0]->ElementsNum();
+    MS_PRINT("Number of tensor elements:%d", tensorNum);
+
+    // Get a pointer to the first score.
+    float *temp_scores = static_cast<float * >(outputTensor[0]->MutableData());
+
+    float scores[RET_CATEGORY_SUM];
+    for (int i = 0; i < RET_CATEGORY_SUM; ++i) {
+        if (temp_scores[i] > 0.5) {
+            MS_PRINT("MindSpore scores[%d] : [%f]", i, temp_scores[i]);
+        }
+        scores[i] = temp_scores[i];
+    }
+
+    // Score for each category.
+    // Converted to text information that needs to be displayed in the APP.
+    std::string categoryScore = "";
+    for (int i = 0; i < RET_CATEGORY_SUM; ++i) {
+        categoryScore += g_labels_name_map[i];
+        categoryScore += ":";
+        std::string score_str = std::to_string(scores[i]);
+        categoryScore += score_str;
+        categoryScore += ";";
+    }
+    return categoryScore;
+}
+
+
+/**
+ * The Java layer reads the model into MappedByteBuffer or ByteBuffer to load the model.
+ */
+extern "C"
+JNIEXPORT jlong JNICALL
+Java_com_huawei_himindsporedemo_gallery_classify_TrackingMobile_loadModel(JNIEnv *env, jobject thiz,
+                                                                          jobject model_buffer,
+                                                                          jint num_thread) {
+    // TODO: implement loadModel()
+    if (nullptr == model_buffer) {
+        MS_PRINT("error, buffer is nullptr!");
+        return (jlong) nullptr;
+    }
+    jlong bufferLen = env->GetDirectBufferCapacity(model_buffer);
+    if (0 == bufferLen) {
+        MS_PRINT("error, bufferLen is 0!");
+        return (jlong) nullptr;
+    }
+
+    char *modelBuffer = CreateLocalModelBuffer(env, model_buffer);
+    if (modelBuffer == nullptr) {
+        MS_PRINT("modelBuffer create failed!");
+        return (jlong) nullptr;
+    }
+
+    // To create a mindspore network inference environment.
+    void **labelEnv = new void *;
+    MSNetWork *labelNet = new MSNetWork;
+    *labelEnv = labelNet;
+
+    lite::Context *context = new lite::Context;
+    context->thread_num_ = num_thread;
+
+    labelNet->CreateSessionMS(modelBuffer, bufferLen, context);
+    delete (context);
+
+    if (labelNet->session == nullptr) {
+        MS_PRINT("MindSpore create session failed!.");
+        return (jlong) nullptr;
+    }
+
+    if (model_buffer != nullptr) {
+        env->DeleteLocalRef(model_buffer);
+    }
+
+    return (jlong) labelEnv;
+}
+
+/**
+ * After the inference environment is successfully created,
+ * sending a picture to the model and run inference.
+ */
+extern "C" JNIEXPORT jstring JNICALL
+Java_com_huawei_himindsporedemo_gallery_classify_TrackingMobile_runNet(JNIEnv *env, jclass type,
+                                                                       jlong netEnv,
+                                                                       jobject srcBitmap) {
+
+    Mat matImageSrc;
+    BitmapToMat(env, srcBitmap, matImageSrc);
+    Mat matImgPreprocessed = PreProcessImageData(matImageSrc);
+
+
+    ImgDims inputDims;
+    inputDims.channel = matImgPreprocessed.channels();
+    inputDims.width = matImgPreprocessed.cols;
+    inputDims.height = matImgPreprocessed.rows;
+
+    // Get the mindsore inference environment which created in loadModel().
+    void **labelEnv = reinterpret_cast<void **>(netEnv);
+    if (labelEnv == nullptr) {
+        MS_PRINT("MindSpore error, labelEnv is a nullptr.");
+        return NULL;
+    }
+    MSNetWork *labelNet = static_cast<MSNetWork *>(*labelEnv);
+
+    auto mSession = labelNet->session;
+    if (mSession == nullptr) {
+        MS_PRINT("MindSpore error, Session is a nullptr.");
+        return NULL;
+    }
+    MS_PRINT("MindSpore get session.");
+
+    auto msInputs = mSession->GetInputs();
+    if (msInputs.size() == 0) {
+        MS_PRINT("MindSpore error, msInputs.size() equals 0.");
+        return NULL;
+    }
+    auto inTensor = msInputs.front();
+
+    // dataHWC is the tensor format.
+    float *dataHWC = new float[inputDims.channel * inputDims.width * inputDims.height];
+    float *ptrTmp = reinterpret_cast<float *>(matImgPreprocessed.data);
+    for (int i = 0; i < inputDims.channel * inputDims.width * inputDims.height; ++i) {
+        dataHWC[i] = ptrTmp[i];
+    }
+
+    // Copy dataHWC to the model input tensor.
+    memcpy(inTensor->MutableData(), dataHWC,
+           inputDims.channel * inputDims.width * inputDims.height * sizeof(float));
+    // When using 'new' to allocate memory space, we need to use 'delete' to free space.
+    delete[] (dataHWC);
+
+    // After the model and image tensor data is loaded, run inference.
+    auto status = mSession->RunGraph();
+
+    if (status != lite::RET_OK) {
+        MS_PRINT("MindSpore run net error.");
+        return NULL;
+    }
+
+    /**
+     * Get the mindspore inference results.
+     * Return the map of output node name and MindSpore Lite MSTensor.
+     */
+    auto msOutputs = mSession->GetOutputMapByNode();
+
+    std::string resultStr = ProcessRunnetResult(msOutputs);
+
+    const char *resultCharData = resultStr.c_str();
+    return (env)->NewStringUTF(resultCharData);
+}
+
+extern "C" JNIEXPORT jboolean JNICALL
+Java_com_huawei_himindsporedemo_gallery_classify_TrackingMobile_unloadModel(JNIEnv *env,
+                                                                            jclass type,
+                                                                            jlong netEnv) {
+    MS_PRINT("MindSpore release net.");
+    void **labelEnv = reinterpret_cast<void **>(netEnv);
+    if (labelEnv == nullptr) {
+        MS_PRINT("MindSpore error, labelEnv is a nullptr.");
+    }
+    MSNetWork *labelNet = static_cast<MSNetWork *>(*labelEnv);
+
+    labelNet->ReleaseNets();
+
+    return (jboolean) true;
+}

model_zoo/official/lite/app/src/main/cpp/MindSporeNetnative.h

@@ -0,0 +1,8 @@
+/*
+ * Copyright (c) Huawei Technologies Co., Ltd. 2018-2019. All rights reserved.
+ */
+
+#ifndef MINDSPORE_JNI_HMS_DEBUG_MINDSPORENETNATIVE_H
+#define MINDSPORE_JNI_HMS_DEBUG_MINDSPORENETNATIVE_H
+
+#endif //MINDSPORE_JNI_HMS_DEBUG_MINDSPORENETNATIVE_H

model_zoo/official/lite/app/src/main/cpp/include/HMS/HMS_label_thres.h

@@ -0,0 +1,616 @@
+/*
+ * Copyright (c) Huawei Technologies Co., Ltd. 2018-2019. All rights reserved.
+ */
+
+#ifndef MNN_JNI_HMS_HMS_LABEL_THRES_H
+#define MNN_JNI_HMS_HMS_LABEL_THRES_H
+
+#include <string.h>
+#include <map>
+
+constexpr int RET_CATEGORY_SUM = 601;
+static std::string g_labels_name_map[RET_CATEGORY_SUM] = {
+    {"Tortoise"},
+    {"Container"},
+    {"Magpie"},
+    {"Seaturtle"},
+    {"Football"},
+    {"Ambulance"},
+    {"Ladder"},
+    {"Toothbrush"},
+    {"Syringe"},
+    {"Sink"},
+    {"Toy"},
+    {"Organ(MusicalInstrument) "},
+    {"Cassettedeck"},
+    {"Apple"},
+    {"Humaneye"},
+    {"Cosmetics"},
+    {"Paddle"},
+    {"Snowman"},
+    {"Beer"},
+    {"Chopsticks"},
+    {"Humanbeard"},
+    {"Bird"},
+    {"Parkingmeter"},
+    {"Trafficlight"},
+    {"Croissant"},
+    {"Cucumber"},
+    {"Radish"},
+    {"Towel"},
+    {"Doll"},
+    {"Skull"},
+    {"Washingmachine"},
+    {"Glove"},
+    {"Tick"},
+    {"Belt"},
+    {"Sunglasses"},
+    {"Banjo"},
+    {"Cart"},
+    {"Ball"},
+    {"Backpack"},
+    {"Bicycle"},
+    {"Homeappliance"},
+    {"Centipede"},
+    {"Boat"},
+    {"Surfboard"},
+    {"Boot"},
+    {"Headphones"},
+    {"Hotdog"},
+    {"Shorts"},
+    {"Fastfood"},
+    {"Bus"},
+    {"Boy "},
+    {"Screwdriver"},
+    {"Bicyclewheel"},
+    {"Barge"},
+    {"Laptop"},
+    {"Miniskirt"},
+    {"Drill(Tool)"},
+    {"Dress"},
+    {"Bear"},
+    {"Waffle"},
+    {"Pancake"},
+    {"Brownbear"},
+    {"Woodpecker"},
+    {"Bluejay"},
+    {"Pretzel"},
+    {"Bagel"},
+    {"Tower"},
+    {"Teapot"},
+    {"Person"},
+    {"Bowandarrow"},
+    {"Swimwear"},
+    {"Beehive"},
+    {"Brassiere"},
+    {"Bee"},
+    {"Bat(Animal)"},
+    {"Starfish"},
+    {"Popcorn"},
+    {"Burrito"},
+    {"Chainsaw"},
+    {"Balloon"},
+    {"Wrench"},
+    {"Tent"},
+    {"Vehicleregistrationplate"},
+    {"Lantern"},
+    {"Toaster"},
+    {"Flashlight"},
+    {"Billboard"},
+    {"Tiara"},
+    {"Limousine"},
+    {"Necklace"},
+    {"Carnivore"},
+    {"Scissors"},
+    {"Stairs"},
+    {"Computerkeyboard"},
+    {"Printer"},
+    {"Trafficsign"},
+    {"Chair"},
+    {"Shirt"},
+    {"Poster"},
+    {"Cheese"},
+    {"Sock"},
+    {"Firehydrant"},
+    {"Landvehicle"},
+    {"Earrings"},
+    {"Tie"},
+    {"Watercraft"},
+    {"Cabinetry"},
+    {"Suitcase"},
+    {"Muffin"},
+    {"Bidet"},
+    {"Snack"},
+    {"Snowmobile"},
+    {"Clock"},
+    {"Medicalequipment"},
+    {"Cattle"},
+    {"Cello"},
+    {"Jetski"},
+    {"Camel"},
+    {"Coat"},
+    {"Suit"},
+    {"Desk"},
+    {"Cat"},
+    {"Bronzesculpture"},
+    {"Juice"},
+    {"Gondola"},
+    {"Beetle"},
+    {"Cannon"},
+    {"Computermouse"},
+    {"Cookie"},
+    {"Officebuilding"},
+    {"Fountain"},
+    {"Coin"},
+    {"Calculator"},
+    {"Cocktail"},
+    {"Computermonitor"},
+    {"Box"},
+    {"Stapler"},
+    {"Christmastree"},
+    {"Cowboyhat"},
+    {"Hikingequipment"},
+    {"Studiocouch"},
+    {"Drum"},
+    {"Dessert"},
+    {"Winerack"},
+    {"Drink"},
+    {"Zucchini"},
+    {"Ladle"},
+    {"Humanmouth"},
+    {"DairyProduct"},
+    {"Dice"},
+    {"Oven"},
+    {"Dinosaur"},
+    {"Ratchet(Device)"},
+    {"Couch"},
+    {"Cricketball"},
+    {"Wintermelon"},
+    {"Spatula"},
+    {"Whiteboard"},
+    {"Pencilsharpener"},
+    {"Door"},
+    {"Hat"},
+    {"Shower"},
+    {"Eraser"},
+    {"Fedora"},
+    {"Guacamole"},
+    {"Dagger"},
+    {"Scarf"},
+    {"Dolphin"},
+    {"Sombrero"},
+    {"Tincan"},
+    {"Mug"},
+    {"Tap"},
+    {"Harborseal"},
+    {"Stretcher"},
+    {"Canopener"},
+    {"Goggles"},
+    {"Humanbody"},
+    {"Rollerskates"},
+    {"Coffeecup"},
+    {"Cuttingboard"},
+    {"Blender"},
+    {"Plumbingfixture"},
+    {"Stopsign"},
+    {"Officesupplies"},
+    {"Volleyball(Ball)"},
+    {"Vase"},
+    {"Slowcooker"},
+    {"Wardrobe"},
+    {"Coffee"},
+    {"Whisk"},
+    {"Papertowel"},
+    {"Personalcare"},
+    {"Food"},
+    {"Sunhat"},
+    {"Treehouse"},
+    {"Flyingdisc"},
+    {"Skirt"},
+    {"Gasstove"},
+    {"Saltandpeppershakers"},
+    {"Mechanicalfan"},
+    {"Facepowder"},
+    {"Fax"},
+    {"Fruit"},
+    {"Frenchfries"},
+    {"Nightstand"},
+    {"Barrel"},
+    {"Kite"},
+    {"Tart"},
+    {"Treadmill"},
+    {"Fox"},
+    {"Flag"},
+    {"Frenchhorn"},
+    {"Windowblind"},
+    {"Humanfoot"},
+    {"Golfcart"},
+    {"Jacket"},
+    {"Egg(Food)"},
+    {"Streetlight"},
+    {"Guitar"},
+    {"Pillow"},
+    {"Humanleg"},
+    {"Isopod"},
+    {"Grape"},
+    {"Humanear"},
+    {"Powerplugsandsockets"},
+    {"Panda"},
+    {"Giraffe"},
+    {"Woman"},
+    {"Doorhandle"},
+    {"Rhinoceros"},
+    {"Bathtub"},
+    {"Goldfish"},
+    {"Houseplant"},
+    {"Goat"},
+    {"Baseballbat"},
+    {"Baseballglove"},
+    {"Mixingbowl"},
+    {"Marineinvertebrates"},
+    {"Kitchenutensil"},
+    {"Lightswitch"},
+    {"House"},
+    {"Horse"},
+    {"Stationarybicycle"},
+    {"Hammer"},
+    {"Ceilingfan"},
+    {"Sofabed"},
+    {"Adhesivetape "},
+    {"Harp"},
+    {"Sandal"},
+    {"Bicyclehelmet"},
+    {"Saucer"},
+    {"Harpsichord"},
+    {"Humanhair"},
+    {"Heater"},
+    {"Harmonica"},
+    {"Hamster"},
+    {"Curtain"},
+    {"Bed"},
+    {"Kettle"},
+    {"Fireplace"},
+    {"Scale"},
+    {"Drinkingstraw"},
+    {"Insect"},
+    {"Hairdryer"},
+    {"Kitchenware"},
+    {"Indoorrower"},
+    {"Invertebrate"},
+    {"Foodprocessor"},
+    {"Bookcase"},
+    {"Refrigerator"},
+    {"Wood-burningstove"},
+    {"Punchingbag"},
+    {"Commonfig"},
+    {"Cocktailshaker"},
+    {"Jaguar(Animal)"},
+    {"Golfball"},
+    {"Fashionaccessory"},
+    {"Alarmclock"},
+    {"Filingcabinet"},
+    {"Artichoke"},
+    {"Table"},
+    {"Tableware"},
+    {"Kangaroo"},
+    {"Koala"},
+    {"Knife"},
+    {"Bottle"},
+    {"Bottleopener"},
+    {"Lynx"},
+    {"Lavender(Plant)"},
+    {"Lighthouse"},
+    {"Dumbbell"},
+    {"Humanhead"},
+    {"Bowl"},
+    {"Humidifier"},
+    {"Porch"},
+    {"Lizard"},
+    {"Billiardtable"},
+    {"Mammal"},
+    {"Mouse"},
+    {"Motorcycle"},
+    {"Musicalinstrument"},
+    {"Swimcap"},
+    {"Fryingpan"},
+    {"Snowplow"},
+    {"Bathroomcabinet"},
+    {"Missile"},
+    {"Bust"},
+    {"Man"},
+    {"Waffleiron"},
+    {"Milk"},
+    {"Ringbinder"},
+    {"Plate"},
+    {"Mobilephone"},
+    {"Bakedgoods"},
+    {"Mushroom"},
+    {"Crutch"},
+    {"Pitcher(Container)"},
+    {"Mirror"},
+    {"Personalflotationdevice"},
+    {"Tabletennisracket"},
+    {"Pencilcase"},
+    {"Musicalkeyboard"},
+    {"Scoreboard"},
+    {"Briefcase"},
+    {"Kitchenknife"},
+    {"Nail(Construction)"},
+    {"Tennisball"},
+    {"Plasticbag"},
+    {"Oboe"},
+    {"Chestofdrawers"},
+    {"Ostrich"},
+    {"Piano"},
+    {"Girl"},
+    {"Plant"},
+    {"Potato"},
+    {"Hairspray"},
+    {"Sportsequipment"},
+    {"Pasta"},
+    {"Penguin"},
+    {"Pumpkin"},
+    {"Pear"},
+    {"Infantbed"},
+    {"Polarbear"},
+    {"Mixer"},
+    {"Cupboard"},
+    {"Jacuzzi"},
+    {"Pizza"},
+    {"Digitalclock"},
+    {"Pig"},
+    {"Reptile"},
+    {"Rifle"},
+    {"Lipstick"},
+    {"Skateboard"},
+    {"Raven"},
+    {"Highheels"},
+    {"Redpanda"},
+    {"Rose"},
+    {"Rabbit"},
+    {"Sculpture"},
+    {"Saxophone"},
+    {"Shotgun"},
+    {"Seafood"},
+    {"Submarinesandwich"},
+    {"Snowboard"},
+    {"Sword"},
+    {"Pictureframe"},
+    {"Sushi"},
+    {"Loveseat"},
+    {"Ski"},
+    {"Squirrel"},
+    {"Tripod"},
+    {"Stethoscope"},
+    {"Submarine"},
+    {"Scorpion"},
+    {"Segway"},
+    {"Trainingbench"},
+    {"Snake"},
+    {"Coffeetable"},
+    {"Skyscraper"},
+    {"Sheep"},
+    {"Television"},
+    {"Trombone"},
+    {"Tea"},
+    {"Tank"},
+    {"Taco"},
+    {"Telephone"},
+    {"Torch"},
+    {"Tiger"},
+    {"Strawberry"},
+    {"Trumpet"},
+    {"Tree"},
+    {"Tomato"},
+    {"Train"},
+    {"Tool"},
+    {"Picnicbasket"},
+    {"Cookingspray"},
+    {"Trousers"},
+    {"Bowlingequipment"},
+    {"Footballhelmet"},
+    {"Truck"},
+    {"Measuringcup"},
+    {"Coffeemaker"},
+    {"Violin"},
+    {"Vehicle"},
+    {"Handbag"},
+    {"Papercutter"},
+    {"Wine"},
+    {"Weapon"},
+    {"Wheel"},
+    {"Worm"},
+    {"Wok"},
+    {"Whale"},
+    {"Zebra"},
+    {"Autopart"},
+    {"Jug"},
+    {"Pizzacutter"},
+    {"Cream"},
+    {"Monkey"},
+    {"Lion"},
+    {"Bread"},
+    {"Platter"},
+    {"Chicken"},
+    {"Eagle"},
+    {"Helicopter"},
+    {"Owl"},
+    {"Duck"},
+    {"Turtle"},
+    {"Hippopotamus"},
+    {"Crocodile"},
+    {"Toilet"},
+    {"Toiletpaper"},
+    {"Squid"},
+    {"Clothing"},
+    {"Footwear"},
+    {"Lemon"},
+    {"Spider"},
+    {"Deer"},
+    {"Frog"},
+    {"Banana"},
+    {"Rocket"},
+    {"Wineglass"},
+    {"Countertop"},
+    {"Tabletcomputer"},
+    {"Wastecontainer"},
+    {"Swimmingpool"},
+    {"Dog"},
+    {"Book"},
+    {"Elephant"},
+    {"Shark"},
+    {"Candle"},
+    {"Leopard"},
+    {"Axe"},
+    {"Handdryer"},
+    {"Soapdispenser"},
+    {"Porcupine"},
+    {"Flower"},
+    {"Canary"},
+    {"Cheetah"},
+    {"Palmtree"},
+    {"Hamburger"},
+    {"Maple"},
+    {"Building"},
+    {"Fish"},
+    {"Lobster"},
+    {"GardenAsparagus"},
+    {"Furniture"},
+    {"Hedgehog"},
+    {"Airplane"},
+    {"Spoon"},
+    {"Otter"},
+    {"Bull"},
+    {"Oyster"},
+    {"Horizontalbar"},
+    {"Conveniencestore"},
+    {"Bomb"},
+    {"Bench"},
+    {"Icecream"},
+    {"Caterpillar"},
+    {"Butterfly"},
+    {"Parachute"},
+    {"Orange"},
+    {"Antelope"},
+    {"Beaker"},
+    {"Mothsandbutterflies"},
+    {"Window"},
+    {"Closet"},
+    {"Castle"},
+    {"Jellyfish"},
+    {"Goose"},
+    {"Mule"},
+    {"Swan"},
+    {"Peach"},
+    {"Coconut"},
+    {"Seatbelt"},
+    {"Raccoon"},
+    {"Chisel"},
+    {"Fork"},
+    {"Lamp"},
+    {"Camera"},
+    {"Squash(Plant)"},
+    {"Racket"},
+    {"Humanface"},
+    {"Humanarm"},
+    {"Vegetable"},
+    {"Diaper"},
+    {"Unicycle"},
+    {"Falcon"},
+    {"Chime"},
+    {"Snail"},
+    {"Shellfish"},
+    {"Cabbage"},
+    {"Carrot"},
+    {"Mango"},
+    {"Jeans"},
+    {"Flowerpot"},
+    {"Pineapple"},
+    {"Drawer"},
+    {"Stool"},
+    {"Envelope"},
+    {"Cake"},
+    {"Dragonfly"},
+    {"Commonsunflower"},
+    {"Microwaveoven"},
+    {"Honeycomb"},
+    {"Marinemammal"},
+    {"Sealion"},
+    {"Ladybug"},
+    {"Shelf"},
+    {"Watch"},
+    {"Candy"},
+    {"Salad"},
+    {"Parrot"},
+    {"Handgun"},
+    {"Sparrow"},
+    {"Van"},
+    {"Grinder"},
+    {"Spicerack"},
+    {"Lightbulb"},
+    {"Cordedphone"},
+    {"Sportsuniform"},
+    {"Tennisracket"},
+    {"Wallclock"},
+    {"Servingtray"},
+    {"Kitchen&diningroomtable"},
+    {"Dogbed"},
+    {"Cakestand"},
+    {"Catfurniture"},
+    {"Bathroomaccessory"},
+    {"Facialtissueholder"},
+    {"Pressurecooker"},
+    {"Kitchenappliance"},
+    {"Tire"},
+    {"Ruler"},
+    {"Luggageandbags"},
+    {"Microphone"},
+    {"Broccoli"},
+    {"Umbrella"},
+    {"Pastry"},
+    {"Grapefruit"},
+    {"Band-aid"},
+    {"Animal"},
+    {"Bellpepper"},
+    {"Turkey"},
+    {"Lily"},
+    {"Pomegranate"},
+    {"Doughnut"},
+    {"Glasses"},
+    {"Humannose"},
+    {"Pen"},
+    {"Ant"},
+    {"Car"},
+    {"Aircraft"},
+    {"Humanhand"},
+    {"Skunk"},
+    {"Teddybear"},
+    {"Watermelon"},
+    {"Cantaloupe"},
+    {"Dishwasher"},
+    {"Flute"},
+    {"Balancebeam"},
+    {"Sandwich"},
+    {"Shrimp"},
+    {"Sewingmachine"},
+    {"Binoculars"},
+    {"Raysandskates"},
+    {"Ipod"},
+    {"Accordion"},
+    {"Willow"},
+    {"Crab"},
+    {"Crown"},
+    {"Seahorse"},
+    {"Perfume"},
+    {"Alpaca"},
+    {"Taxi"},
+    {"Canoe"},
+    {"Remotecontrol"},
+    {"Wheelchair"},
+    {"Rugbyball"},
+    {"Armadillo"},
+    {"Maracas"},
+    {"Helmet"},
+};
+
+#endif // MNN_JNI_HMS_HMS_LABEL_THRES_H

model_zoo/official/lite/app/src/main/cpp/include/MindSpore/MSNetWork.cpp

@@ -0,0 +1,41 @@
+#include "MSNetWork.h"
+#include <iostream>
+#include <android/log.h>
+#include "errorcode.h"
+
+#define MS_PRINT(format, ...) __android_log_print(ANDROID_LOG_INFO, "MSJNI", format, ##__VA_ARGS__)
+
+MSNetWork::MSNetWork(void) : session(nullptr) {}
+MSNetWork::~MSNetWork(void) {}
+
+
+void MSNetWork::CreateSessionMS(char* modelBuffer, size_t bufferLen, mindspore::lite::Context* ctx)
+{
+    session = mindspore::session::LiteSession::CreateSession(ctx);
+    if (session == nullptr){
+        MS_PRINT("Create Session failed.");
+        return;
+    }
+
+    // Compile model.
+    auto model = mindspore::lite::Model::Import(modelBuffer, bufferLen);
+    if (model == nullptr){
+        MS_PRINT("Import model failed.");
+        return;
+    }
+
+    int ret = session->CompileGraph(model);
+    if (ret != mindspore::lite::RET_OK){
+        MS_PRINT("CompileGraph failed.");
+        return;
+    }
+
+}
+
+int MSNetWork::ReleaseNets(void)
+{
+    delete session;
+//    delete model;
+    return 0;
+}
+

model_zoo/official/lite/app/src/main/cpp/include/MindSpore/MSNetWork.h

@@ -0,0 +1,50 @@
+// * Copyright (c) Huawei Technologies Co., Ltd. 2018-2019. All rights reserved.
+
+#ifndef MSNETWORK_H
+#define MSNETWORK_H
+
+#include <cstdio>
+#include <algorithm>
+#include <fstream>
+#include <functional>
+#include <sstream>
+#include <vector>
+#include <map>
+#include <string>
+#include <memory>
+#include <utility>
+
+#include <context.h>
+#include <lite_session.h>
+#include <model.h>
+#include <errorcode.h>
+
+using namespace mindspore;
+
+struct ImgDims {
+    int channel = 0;
+    int width = 0;
+    int height = 0;
+};
+
+/*struct SessIterm {
+    std::shared_ptr<mindspore::session::LiteSession> sess = nullptr;
+};*/
+
+
+
+class MSNetWork {
+public:
+    MSNetWork();
+    ~MSNetWork();
+
+    void CreateSessionMS(char* modelBuffer, size_t bufferLen, mindspore::lite::Context* ctx);
+    int ReleaseNets(void);
+    mindspore::session::LiteSession *session;
+    mindspore::lite::Model *model;
+
+private:
+    //std::map<std::string, SessIterm> sess;
+};
+
+#endif

model_zoo/official/lite/app/src/main/cpp/include/MindSpore/context.h

@@ -0,0 +1,71 @@
+/**
+ * Copyright 2020 Huawei Technologies Co., Ltd
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef MINDSPORE_LITE_INCLUDE_CONTEXT_H_
+#define MINDSPORE_LITE_INCLUDE_CONTEXT_H_
+
+#include <string>
+#include <memory>
+#include "ms_tensor.h"
+#include "thread_pool_config.h"
+
+namespace mindspore {
+    namespace lite {
+        /// \brief Allocator defined a memory pool for malloc memory and free memory dynamically.
+        ///
+        /// \note List public class and interface for reference.
+        class Allocator;
+
+        /// \brief DeviceType defined for holding user's preferred backend.
+        typedef enum {
+            DT_CPU, /**< CPU device type */
+            DT_GPU, /**< GPU device type */
+            DT_NPU  /**< NPU device type, not supported yet */
+        } DeviceType;
+
+        /// \brief DeviceContext defined for holding DeviceType.
+        typedef struct {
+            DeviceType type; /**< device type */
+        } DeviceContext;
+
+        /// \brief Context defined for holding environment variables during runtime.
+        class MS_API Context {
+        public:
+            /// \brief Constructor of MindSpore Lite Context using default value for parameters.
+            ///
+            /// \return Instance of MindSpore Lite Context.
+            Context();
+
+            /// \brief Constructor of MindSpore Lite Context using input value for parameters.
+            ///
+            /// \param[in] thread_num Define the work thread number during the runtime.
+            /// \param[in] allocator Define the allocator for malloc.
+            /// \param[in] device_ctx Define device information during the runtime.
+            Context(int thread_num, std::shared_ptr<Allocator> allocator, DeviceContext device_ctx);
+
+            /// \brief Destructor of MindSpore Lite Context.
+            virtual ~Context();
+
+        public:
+            bool float16_priority = false; /**< allow priority select float16 kernel */
+            DeviceContext device_ctx_{DT_CPU};
+            int thread_num_ = 2; /**< thread number config for thread pool */
+            std::shared_ptr<Allocator> allocator = nullptr;
+            CpuBindMode cpu_bind_mode_ = MID_CPU;
+        };
+    }
+}  // namespace mindspore::lite
+#endif  // MINDSPORE_LITE_INCLUDE_CONTEXT_H_

model_zoo/official/lite/app/src/main/cpp/include/MindSpore/errorcode.h

@@ -0,0 +1,59 @@
+/**
+ * Copyright 2020 Huawei Technologies Co., Ltd
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef MINDSPORE_LITE_INCLUDE_ERRORCODE_H_
+#define MINDSPORE_LITE_INCLUDE_ERRORCODE_H_
+
+namespace mindspore {
+namespace lite {
+/// \brief STATUS defined for holding error code in MindSpore Lite.
+using STATUS = int;
+
+/* Success */
+constexpr int RET_OK = 0; /**< No error occurs. */
+
+/* Common error code, range: [-1, -100]*/
+constexpr int RET_ERROR = -1;         /**< Common error code. */
+constexpr int RET_NULL_PTR = -2;      /**< NULL pointer returned.*/
+constexpr int RET_PARAM_INVALID = -3; /**< Invalid parameter.*/
+constexpr int RET_NO_CHANGE = -4;     /**< No change. */
+constexpr int RET_SUCCESS_EXIT = -5;  /**< No error but exit. */
+constexpr int RET_MEMORY_FAILED = -6; /**< Fail to create memory. */
+
+/* Executor error code, range: [-101,-200] */
+constexpr int RET_OUT_OF_TENSOR_RANGE = -101; /**< Failed to check range. */
+constexpr int RET_INPUT_TENSOR_ERROR = -102;  /**< Failed to check input tensor. */
+constexpr int RET_REENTRANT_ERROR = -103;     /**< Exist executor running. */
+
+/* Graph error code, range: [-201,-300] */
+constexpr int RET_GRAPH_FILE_ERR = -201; /**< Failed to verify graph file. */
+
+/* Node error code, range: [-301,-400] */
+constexpr int RET_NOT_FIND_OP = -301;        /**< Failed to find operator. */
+constexpr int RET_INVALID_OP_NAME = -302;    /**< Invalid operator name. */
+constexpr int RET_INVALID_OP_ATTR = -303;    /**< Invalid operator attr. */
+constexpr int RET_OP_EXECUTE_FAILURE = -304; /**< Failed to execution operator. */
+
+/* Tensor error code, range: [-401,-500] */
+constexpr int RET_FORMAT_ERR = -401; /**< Failed to checking tensor format. */
+
+/* InferShape error code, range: [-501,-600] */
+constexpr int RET_INFER_ERR = -501; /**< Failed to infer shape. */
+constexpr int RET_INFER_INVALID = -502; /**< Invalid infer shape before runtime. */
+}  // namespace lite
+}  // namespace mindspore
+
+#endif  // MINDSPORE_LITE_INCLUDE_ERRORCODE_H_

model_zoo/official/lite/app/src/main/cpp/include/MindSpore/flatbuffers/include/base.h

@@ -0,0 +1,379 @@
+#ifndef FLATBUFFERS_BASE_H_
+#define FLATBUFFERS_BASE_H_
+
+// clang-format off
+
+// If activate should be declared and included first.
+#if defined(FLATBUFFERS_MEMORY_LEAK_TRACKING) && \
+    defined(_MSC_VER) && defined(_DEBUG)
+  // The _CRTDBG_MAP_ALLOC inside <crtdbg.h> will replace
+  // calloc/free (etc) to its debug version using #define directives.
+  #define _CRTDBG_MAP_ALLOC
+  #include <stdlib.h>
+  #include <crtdbg.h>
+  // Replace operator new by trace-enabled version.
+  #define DEBUG_NEW new(_NORMAL_BLOCK, __FILE__, __LINE__)
+  #define new DEBUG_NEW
+#endif
+
+#if !defined(FLATBUFFERS_ASSERT)
+#include <assert.h>
+#define FLATBUFFERS_ASSERT assert
+#elif defined(FLATBUFFERS_ASSERT_INCLUDE)
+// Include file with forward declaration
+#include FLATBUFFERS_ASSERT_INCLUDE
+#endif
+
+#ifndef ARDUINO
+#include <cstdint>
+#endif
+
+#include <cstddef>
+#include <cstdlib>
+#include <cstring>
+
+#if defined(ARDUINO) && !defined(ARDUINOSTL_M_H)
+  #include <utility.h>
+#else
+  #include <utility>
+#endif
+
+#include <string>
+#include <type_traits>
+#include <vector>
+#include <set>
+#include <algorithm>
+#include <iterator>
+#include <memory>
+
+#ifdef _STLPORT_VERSION
+  #define FLATBUFFERS_CPP98_STL
+#endif
+#ifndef FLATBUFFERS_CPP98_STL
+  #include <functional>
+#endif
+
+#include "stl_emulation.h"
+
+// Note the __clang__ check is needed, because clang presents itself
+// as an older GNUC compiler (4.2).
+// Clang 3.3 and later implement all of the ISO C++ 2011 standard.
+// Clang 3.4 and later implement all of the ISO C++ 2014 standard.
+// http://clang.llvm.org/cxx_status.html
+
+// Note the MSVC value '__cplusplus' may be incorrect:
+// The '__cplusplus' predefined macro in the MSVC stuck at the value 199711L,
+// indicating (erroneously!) that the compiler conformed to the C++98 Standard.
+// This value should be correct starting from MSVC2017-15.7-Preview-3.
+// The '__cplusplus' will be valid only if MSVC2017-15.7-P3 and the `/Zc:__cplusplus` switch is set.
+// Workaround (for details see MSDN):
+// Use the _MSC_VER and _MSVC_LANG definition instead of the __cplusplus  for compatibility.
+// The _MSVC_LANG macro reports the Standard version regardless of the '/Zc:__cplusplus' switch.
+
+#if defined(__GNUC__) && !defined(__clang__)
+  #define FLATBUFFERS_GCC (__GNUC__ * 10000 + __GNUC_MINOR__ * 100 + __GNUC_PATCHLEVEL__)
+#else
+  #define FLATBUFFERS_GCC 0
+#endif
+
+#if defined(__clang__)
+  #define FLATBUFFERS_CLANG (__clang_major__ * 10000 + __clang_minor__ * 100 + __clang_patchlevel__)
+#else
+  #define FLATBUFFERS_CLANG 0
+#endif
+
+/// @cond FLATBUFFERS_INTERNAL
+#if __cplusplus <= 199711L && \
+    (!defined(_MSC_VER) || _MSC_VER < 1600) && \
+    (!defined(__GNUC__) || \
+      (__GNUC__ * 10000 + __GNUC_MINOR__ * 100 + __GNUC_PATCHLEVEL__ < 40400))
+  #error A C++11 compatible compiler with support for the auto typing is \
+         required for FlatBuffers.
+  #error __cplusplus _MSC_VER __GNUC__  __GNUC_MINOR__  __GNUC_PATCHLEVEL__
+#endif
+
+#if !defined(__clang__) && \
+    defined(__GNUC__) && \
+    (__GNUC__ * 10000 + __GNUC_MINOR__ * 100 + __GNUC_PATCHLEVEL__ < 40600)
+  // Backwards compatability for g++ 4.4, and 4.5 which don't have the nullptr
+  // and constexpr keywords. Note the __clang__ check is needed, because clang
+  // presents itself as an older GNUC compiler.
+  #ifndef nullptr_t
+    const class nullptr_t {
+    public:
+      template<class T> inline operator T*() const { return 0; }
+    private:
+      void operator&() const;
+    } nullptr = {};
+  #endif
+  #ifndef constexpr
+    #define constexpr const
+  #endif
+#endif
+
+// The wire format uses a little endian encoding (since that's efficient for
+// the common platforms).
+#if defined(__s390x__)
+  #define FLATBUFFERS_LITTLEENDIAN 0
+#endif // __s390x__
+#if !defined(FLATBUFFERS_LITTLEENDIAN)
+  #if defined(__GNUC__) || defined(__clang__)
+    #if (defined(__BIG_ENDIAN__) || \
+         (defined(__BYTE_ORDER__) && __BYTE_ORDER__ == __ORDER_BIG_ENDIAN__))
+      #define FLATBUFFERS_LITTLEENDIAN 0
+    #else
+      #define FLATBUFFERS_LITTLEENDIAN 1
+    #endif // __BIG_ENDIAN__
+  #elif defined(_MSC_VER)
+    #if defined(_M_PPC)
+      #define FLATBUFFERS_LITTLEENDIAN 0
+    #else
+      #define FLATBUFFERS_LITTLEENDIAN 1
+    #endif
+  #else
+    #error Unable to determine endianness, define FLATBUFFERS_LITTLEENDIAN.
+  #endif
+#endif // !defined(FLATBUFFERS_LITTLEENDIAN)
+
+#define FLATBUFFERS_VERSION_MAJOR 1
+#define FLATBUFFERS_VERSION_MINOR 11
+#define FLATBUFFERS_VERSION_REVISION 0
+#define FLATBUFFERS_STRING_EXPAND(X) #X
+#define FLATBUFFERS_STRING(X) FLATBUFFERS_STRING_EXPAND(X)
+
+#if (!defined(_MSC_VER) || _MSC_VER > 1600) && \
+    (!defined(__GNUC__) || (__GNUC__ * 100 + __GNUC_MINOR__ >= 407)) || \
+    defined(__clang__)
+  #define FLATBUFFERS_FINAL_CLASS final
+  #define FLATBUFFERS_OVERRIDE override
+  #define FLATBUFFERS_VTABLE_UNDERLYING_TYPE : flatbuffers::voffset_t
+#else
+  #define FLATBUFFERS_FINAL_CLASS
+  #define FLATBUFFERS_OVERRIDE
+  #define FLATBUFFERS_VTABLE_UNDERLYING_TYPE
+#endif
+
+#if (!defined(_MSC_VER) || _MSC_VER >= 1900) && \
+    (!defined(__GNUC__) || (__GNUC__ * 100 + __GNUC_MINOR__ >= 406)) || \
+    (defined(__cpp_constexpr) && __cpp_constexpr >= 200704)
+  #define FLATBUFFERS_CONSTEXPR constexpr
+#else
+  #define FLATBUFFERS_CONSTEXPR const
+#endif
+
+#if (defined(__cplusplus) && __cplusplus >= 201402L) || \
+    (defined(__cpp_constexpr) && __cpp_constexpr >= 201304)
+  #define FLATBUFFERS_CONSTEXPR_CPP14 FLATBUFFERS_CONSTEXPR
+#else
+  #define FLATBUFFERS_CONSTEXPR_CPP14
+#endif
+
+#if (defined(__GXX_EXPERIMENTAL_CXX0X__) && (__GNUC__ * 100 + __GNUC_MINOR__ >= 406)) || \
+    (defined(_MSC_FULL_VER) && (_MSC_FULL_VER >= 190023026)) || \
+    defined(__clang__)
+  #define FLATBUFFERS_NOEXCEPT noexcept
+#else
+  #define FLATBUFFERS_NOEXCEPT
+#endif
+
+// NOTE: the FLATBUFFERS_DELETE_FUNC macro may change the access mode to
+// private, so be sure to put it at the end or reset access mode explicitly.
+#if (!defined(_MSC_VER) || _MSC_FULL_VER >= 180020827) && \
+    (!defined(__GNUC__) || (__GNUC__ * 100 + __GNUC_MINOR__ >= 404)) || \
+    defined(__clang__)
+  #define FLATBUFFERS_DELETE_FUNC(func) func = delete;
+#else
+  #define FLATBUFFERS_DELETE_FUNC(func) private: func;
+#endif
+
+#ifndef FLATBUFFERS_HAS_STRING_VIEW
+  // Only provide flatbuffers::string_view if __has_include can be used
+  // to detect a header that provides an implementation
+  #if defined(__has_include)
+    // Check for std::string_view (in c++17)
+    #if __has_include(<string_view>) && (__cplusplus >= 201606 || _HAS_CXX17)
+      #include <string_view>
+      namespace flatbuffers {
+        typedef std::string_view string_view;
+      }
+      #define FLATBUFFERS_HAS_STRING_VIEW 1
+    // Check for std::experimental::string_view (in c++14, compiler-dependent)
+    #elif __has_include(<experimental/string_view>) && (__cplusplus >= 201411)
+      #include <experimental/string_view>
+      namespace flatbuffers {
+        typedef std::experimental::string_view string_view;
+      }
+      #define FLATBUFFERS_HAS_STRING_VIEW 1
+    #endif
+  #endif // __has_include
+#endif // !FLATBUFFERS_HAS_STRING_VIEW
+
+#ifndef FLATBUFFERS_HAS_NEW_STRTOD
+  // Modern (C++11) strtod and strtof functions are available for use.
+  // 1) nan/inf strings as argument of strtod;
+  // 2) hex-float  as argument of  strtod/strtof.
+  #if (defined(_MSC_VER) && _MSC_VER >= 1900) || \
+      (defined(__GNUC__) && (__GNUC__ * 100 + __GNUC_MINOR__ >= 409)) || \
+      (defined(__clang__))
+    #define FLATBUFFERS_HAS_NEW_STRTOD 1
+  #endif
+#endif // !FLATBUFFERS_HAS_NEW_STRTOD
+
+#ifndef FLATBUFFERS_LOCALE_INDEPENDENT
+  // Enable locale independent functions {strtof_l, strtod_l,strtoll_l, strtoull_l}.
+  // They are part of the POSIX-2008 but not part of the C/C++ standard.
+  // GCC/Clang have definition (_XOPEN_SOURCE>=700) if POSIX-2008.
+  #if ((defined(_MSC_VER) && _MSC_VER >= 1800)            || \
+       (defined(_XOPEN_SOURCE) && (_XOPEN_SOURCE>=700)))
+    #define FLATBUFFERS_LOCALE_INDEPENDENT 1
+  #else
+    #define FLATBUFFERS_LOCALE_INDEPENDENT 0
+  #endif
+#endif  // !FLATBUFFERS_LOCALE_INDEPENDENT
+
+// Suppress Undefined Behavior Sanitizer (recoverable only). Usage:
+// - __supress_ubsan__("undefined")
+// - __supress_ubsan__("signed-integer-overflow")
+#if defined(__clang__)
+  #define __supress_ubsan__(type) __attribute__((no_sanitize(type)))
+#elif defined(__GNUC__) && (__GNUC__ * 100 + __GNUC_MINOR__ >= 409)
+  #define __supress_ubsan__(type) __attribute__((no_sanitize_undefined))
+#else
+  #define __supress_ubsan__(type)
+#endif
+
+// This is constexpr function used for checking compile-time constants.
+// Avoid `#pragma warning(disable: 4127) // C4127: expression is constant`.
+template<typename T> FLATBUFFERS_CONSTEXPR inline bool IsConstTrue(T t) {
+  return !!t;
+}
+
+// Enable C++ attribute [[]] if std:c++17 or higher.
+#if ((__cplusplus >= 201703L) \
+    || (defined(_MSVC_LANG) &&  (_MSVC_LANG >= 201703L)))
+  // All attributes unknown to an implementation are ignored without causing an error.
+  #define FLATBUFFERS_ATTRIBUTE(attr) [[attr]]
+
+  #define FLATBUFFERS_FALLTHROUGH() [[fallthrough]]
+#else
+  #define FLATBUFFERS_ATTRIBUTE(attr)
+
+  #if FLATBUFFERS_CLANG >= 30800
+    #define FLATBUFFERS_FALLTHROUGH() [[clang::fallthrough]]
+  #elif FLATBUFFERS_GCC >= 70300
+    #define FLATBUFFERS_FALLTHROUGH() [[gnu::fallthrough]]
+  #else
+    #define FLATBUFFERS_FALLTHROUGH()
+  #endif
+#endif
+
+/// @endcond
+
+/// @file
+namespace flatbuffers {
+
+/// @cond FLATBUFFERS_INTERNAL
+// Our default offset / size type, 32bit on purpose on 64bit systems.
+// Also, using a consistent offset type maintains compatibility of serialized
+// offset values between 32bit and 64bit systems.
+typedef uint32_t uoffset_t;
+
+// Signed offsets for references that can go in both directions.
+typedef int32_t soffset_t;
+
+// Offset/index used in v-tables, can be changed to uint8_t in
+// format forks to save a bit of space if desired.
+typedef uint16_t voffset_t;
+
+typedef uintmax_t largest_scalar_t;
+
+// In 32bits, this evaluates to 2GB - 1
+#define FLATBUFFERS_MAX_BUFFER_SIZE ((1ULL << (sizeof(soffset_t) * 8 - 1)) - 1)
+
+// We support aligning the contents of buffers up to this size.
+#define FLATBUFFERS_MAX_ALIGNMENT 16
+
+#if defined(_MSC_VER)
+  #pragma warning(push)
+  #pragma warning(disable: 4127) // C4127: conditional expression is constant
+#endif
+
+template<typename T> T EndianSwap(T t) {
+  #if defined(_MSC_VER)
+    #define FLATBUFFERS_BYTESWAP16 _byteswap_ushort
+    #define FLATBUFFERS_BYTESWAP32 _byteswap_ulong
+    #define FLATBUFFERS_BYTESWAP64 _byteswap_uint64
+  #else
+    #if defined(__GNUC__) && __GNUC__ * 100 + __GNUC_MINOR__ < 408 && !defined(__clang__)
+      // __builtin_bswap16 was missing prior to GCC 4.8.
+      #define FLATBUFFERS_BYTESWAP16(x) \
+        static_cast<uint16_t>(__builtin_bswap32(static_cast<uint32_t>(x) << 16))
+    #else
+      #define FLATBUFFERS_BYTESWAP16 __builtin_bswap16
+    #endif
+    #define FLATBUFFERS_BYTESWAP32 __builtin_bswap32
+    #define FLATBUFFERS_BYTESWAP64 __builtin_bswap64
+  #endif
+  if (sizeof(T) == 1) {   // Compile-time if-then's.
+    return t;
+  } else if (sizeof(T) == 2) {
+    union { T t; uint16_t i; } u;
+    u.t = t;
+    u.i = FLATBUFFERS_BYTESWAP16(u.i);
+    return u.t;
+  } else if (sizeof(T) == 4) {
+    union { T t; uint32_t i; } u;
+    u.t = t;
+    u.i = FLATBUFFERS_BYTESWAP32(u.i);
+    return u.t;
+  } else if (sizeof(T) == 8) {
+    union { T t; uint64_t i; } u;
+    u.t = t;
+    u.i = FLATBUFFERS_BYTESWAP64(u.i);
+    return u.t;
+  } else {
+    FLATBUFFERS_ASSERT(0);
+  }
+}
+
+#if defined(_MSC_VER)
+  #pragma warning(pop)
+#endif
+
+
+template<typename T> T EndianScalar(T t) {
+  #if FLATBUFFERS_LITTLEENDIAN
+    return t;
+  #else
+    return EndianSwap(t);
+  #endif
+}
+
+template<typename T>
+// UBSAN: C++ aliasing type rules, see std::bit_cast<> for details.
+__supress_ubsan__("alignment")
+T ReadScalar(const void *p) {
+  return EndianScalar(*reinterpret_cast<const T *>(p));
+}
+
+template<typename T>
+// UBSAN: C++ aliasing type rules, see std::bit_cast<> for details.
+__supress_ubsan__("alignment")
+void WriteScalar(void *p, T t) {
+  *reinterpret_cast<T *>(p) = EndianScalar(t);
+}
+
+template<typename T> struct Offset;
+template<typename T> __supress_ubsan__("alignment") void WriteScalar(void *p, Offset<T> t) {
+  *reinterpret_cast<uoffset_t *>(p) = EndianScalar(t.o);
+}
+
+// Computes how many bytes you'd have to pad to be able to write an
+// "scalar_size" scalar if the buffer had grown to "buf_size" (downwards in
+// memory).
+inline size_t PaddingBytes(size_t buf_size, size_t scalar_size) {
+  return ((~buf_size) + 1) & (scalar_size - 1);
+}
+
+}  // namespace flatbuffers
+#endif  // FLATBUFFERS_BASE_H_

model_zoo/official/lite/app/src/main/cpp/include/MindSpore/flatbuffers/include/code_generators.h

@@ -0,0 +1,203 @@
+/*
+ * Copyright 2014 Google Inc. All rights reserved.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef FLATBUFFERS_CODE_GENERATORS_H_
+#define FLATBUFFERS_CODE_GENERATORS_H_
+
+#include <map>
+#include <sstream>
+#include "flatbuffers/idl.h"
+
+namespace flatbuffers {
+
+// Utility class to assist in generating code through use of text templates.
+//
+// Example code:
+//   CodeWriter code;
+//   code.SetValue("NAME", "Foo");
+//   code += "void {{NAME}}() { printf("%s", "{{NAME}}"); }";
+//   code.SetValue("NAME", "Bar");
+//   code += "void {{NAME}}() { printf("%s", "{{NAME}}"); }";
+//   std::cout << code.ToString() << std::endl;
+//
+// Output:
+//  void Foo() { printf("%s", "Foo"); }
+//  void Bar() { printf("%s", "Bar"); }
+class CodeWriter {
+ public:
+  CodeWriter() {}
+
+  // Clears the current "written" code.
+  void Clear() {
+    stream_.str("");
+    stream_.clear();
+  }
+
+  // Associates a key with a value.  All subsequent calls to operator+=, where
+  // the specified key is contained in {{ and }} delimiters will be replaced by
+  // the given value.
+  void SetValue(const std::string &key, const std::string &value) {
+    value_map_[key] = value;
+  }
+
+  // Appends the given text to the generated code as well as a newline
+  // character.  Any text within {{ and }} delimeters is replaced by values
+  // previously stored in the CodeWriter by calling SetValue above.  The newline
+  // will be suppressed if the text ends with the \\ character.
+  void operator+=(std::string text);
+
+  // Returns the current contents of the CodeWriter as a std::string.
+  std::string ToString() const { return stream_.str(); }
+
+ private:
+  std::map<std::string, std::string> value_map_;
+  std::stringstream stream_;
+};
+
+class BaseGenerator {
+ public:
+  virtual bool generate() = 0;
+
+  static std::string NamespaceDir(const Parser &parser, const std::string &path,
+                                  const Namespace &ns);
+
+ protected:
+  BaseGenerator(const Parser &parser, const std::string &path,
+                const std::string &file_name,
+                const std::string qualifying_start,
+                const std::string qualifying_separator)
+      : parser_(parser),
+        path_(path),
+        file_name_(file_name),
+        qualifying_start_(qualifying_start),
+        qualifying_separator_(qualifying_separator) {}
+  virtual ~BaseGenerator() {}
+
+  // No copy/assign.
+  BaseGenerator &operator=(const BaseGenerator &);
+  BaseGenerator(const BaseGenerator &);
+
+  std::string NamespaceDir(const Namespace &ns) const;
+
+  static const char *FlatBuffersGeneratedWarning();
+
+  static std::string FullNamespace(const char *separator, const Namespace &ns);
+
+  static std::string LastNamespacePart(const Namespace &ns);
+
+  // tracks the current namespace for early exit in WrapInNameSpace
+  // c++, java and csharp returns a different namespace from
+  // the following default (no early exit, always fully qualify),
+  // which works for js and php
+  virtual const Namespace *CurrentNameSpace() const { return nullptr; }
+
+  // Ensure that a type is prefixed with its namespace whenever it is used
+  // outside of its namespace.
+  std::string WrapInNameSpace(const Namespace *ns,
+                              const std::string &name) const;
+
+  std::string WrapInNameSpace(const Definition &def) const;
+
+  std::string GetNameSpace(const Definition &def) const;
+
+  const Parser &parser_;
+  const std::string &path_;
+  const std::string &file_name_;
+  const std::string qualifying_start_;
+  const std::string qualifying_separator_;
+};
+
+struct CommentConfig {
+  const char *first_line;
+  const char *content_line_prefix;
+  const char *last_line;
+};
+
+extern void GenComment(const std::vector<std::string> &dc,
+                       std::string *code_ptr, const CommentConfig *config,
+                       const char *prefix = "");
+
+class FloatConstantGenerator {
+ public:
+  virtual ~FloatConstantGenerator() {}
+  std::string GenFloatConstant(const FieldDef &field) const;
+
+ private:
+  virtual std::string Value(double v, const std::string &src) const = 0;
+  virtual std::string Inf(double v) const = 0;
+  virtual std::string NaN(double v) const = 0;
+
+  virtual std::string Value(float v, const std::string &src) const = 0;
+  virtual std::string Inf(float v) const = 0;
+  virtual std::string NaN(float v) const = 0;
+
+  template<typename T>
+  std::string GenFloatConstantImpl(const FieldDef &field) const;
+};
+
+class SimpleFloatConstantGenerator : public FloatConstantGenerator {
+ public:
+  SimpleFloatConstantGenerator(const char *nan_number,
+                               const char *pos_inf_number,
+                               const char *neg_inf_number);
+
+ private:
+  std::string Value(double v,
+                    const std::string &src) const FLATBUFFERS_OVERRIDE;
+  std::string Inf(double v) const FLATBUFFERS_OVERRIDE;
+  std::string NaN(double v) const FLATBUFFERS_OVERRIDE;
+
+  std::string Value(float v, const std::string &src) const FLATBUFFERS_OVERRIDE;
+  std::string Inf(float v) const FLATBUFFERS_OVERRIDE;
+  std::string NaN(float v) const FLATBUFFERS_OVERRIDE;
+
+  const std::string nan_number_;
+  const std::string pos_inf_number_;
+  const std::string neg_inf_number_;
+};
+
+// C++, C#, Java like generator.
+class TypedFloatConstantGenerator : public FloatConstantGenerator {
+ public:
+  TypedFloatConstantGenerator(const char *double_prefix,
+                              const char *single_prefix, const char *nan_number,
+                              const char *pos_inf_number,
+                              const char *neg_inf_number = "");
+
+ private:
+  std::string Value(double v,
+                    const std::string &src) const FLATBUFFERS_OVERRIDE;
+  std::string Inf(double v) const FLATBUFFERS_OVERRIDE;
+
+  std::string NaN(double v) const FLATBUFFERS_OVERRIDE;
+
+  std::string Value(float v, const std::string &src) const FLATBUFFERS_OVERRIDE;
+  std::string Inf(float v) const FLATBUFFERS_OVERRIDE;
+  std::string NaN(float v) const FLATBUFFERS_OVERRIDE;
+
+  std::string MakeNaN(const std::string &prefix) const;
+  std::string MakeInf(bool neg, const std::string &prefix) const;
+
+  const std::string double_prefix_;
+  const std::string single_prefix_;
+  const std::string nan_number_;
+  const std::string pos_inf_number_;
+  const std::string neg_inf_number_;
+};
+
+}  // namespace MindSpore.flatbuffers
+
+#endif  // FLATBUFFERS_CODE_GENERATORS_H_

model_zoo/official/lite/app/src/main/cpp/include/MindSpore/flatbuffers/include/flatc.h

@@ -0,0 +1,96 @@
+/*
+ * Copyright 2017 Google Inc. All rights reserved.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#include <functional>
+#include <limits>
+#include <string>
+#include "flatbuffers/flatbuffers.h"
+#include "flatbuffers/idl.h"
+#include "flatbuffers/util.h"
+
+#ifndef FLATC_H_
+#  define FLATC_H_
+
+namespace flatbuffers {
+
+class FlatCompiler {
+ public:
+  // Output generator for the various programming languages and formats we
+  // support.
+  struct Generator {
+    typedef bool (*GenerateFn)(const flatbuffers::Parser &parser,
+                               const std::string &path,
+                               const std::string &file_name);
+    typedef std::string (*MakeRuleFn)(const flatbuffers::Parser &parser,
+                                      const std::string &path,
+                                      const std::string &file_name);
+
+    GenerateFn generate;
+    const char *generator_opt_short;
+    const char *generator_opt_long;
+    const char *lang_name;
+    bool schema_only;
+    GenerateFn generateGRPC;
+    flatbuffers::IDLOptions::Language lang;
+    const char *generator_help;
+    MakeRuleFn make_rule;
+  };
+
+  typedef void (*WarnFn)(const FlatCompiler *flatc, const std::string &warn,
+                         bool show_exe_name);
+
+  typedef void (*ErrorFn)(const FlatCompiler *flatc, const std::string &err,
+                          bool usage, bool show_exe_name);
+
+  // Parameters required to initialize the FlatCompiler.
+  struct InitParams {
+    InitParams()
+        : generators(nullptr),
+          num_generators(0),
+          warn_fn(nullptr),
+          error_fn(nullptr) {}
+
+    const Generator *generators;
+    size_t num_generators;
+    WarnFn warn_fn;
+    ErrorFn error_fn;
+  };
+
+  explicit FlatCompiler(const InitParams &params) : params_(params) {}
+
+  int Compile(int argc, const char **argv);
+
+  std::string GetUsageString(const char *program_name) const;
+
+ private:
+  void ParseFile(flatbuffers::Parser &parser, const std::string &filename,
+                 const std::string &contents,
+                 std::vector<const char *> &include_directories) const;
+
+  void LoadBinarySchema(Parser &parser, const std::string &filename,
+                        const std::string &contents);
+
+  void Warn(const std::string &warn, bool show_exe_name = true) const;
+
+  void Error(const std::string &err, bool usage = true,
+             bool show_exe_name = true) const;
+
+  InitParams params_;
+};
+
+}  // namespace MindSpore.flatbuffers
+
+#endif  // FLATC_H_

model_zoo/official/lite/app/src/main/cpp/include/MindSpore/flatbuffers/include/grpc.h

@@ -0,0 +1,328 @@
+/*
+ * Copyright 2014 Google Inc. All rights reserved.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef FLATBUFFERS_GRPC_H_
+#define FLATBUFFERS_GRPC_H_
+
+// Helper functionality to glue FlatBuffers and GRPC.
+
+#include "flatbuffers/flatbuffers.h"
+#include "grpc++/support/byte_buffer.h"
+#include "grpc/byte_buffer_reader.h"
+
+namespace flatbuffers {
+namespace grpc {
+
+// Message is a typed wrapper around a buffer that manages the underlying
+// `grpc_slice` and also provides MindSpore.flatbuffers-specific helpers such as `Verify`
+// and `GetRoot`. Since it is backed by a `grpc_slice`, the underlying buffer
+// is refcounted and ownership is be managed automatically.
+template<class T> class Message {
+ public:
+  Message() : slice_(grpc_empty_slice()) {}
+
+  Message(grpc_slice slice, bool add_ref)
+      : slice_(add_ref ? grpc_slice_ref(slice) : slice) {}
+
+  Message &operator=(const Message &other) = delete;
+
+  Message(Message &&other) : slice_(other.slice_) {
+    other.slice_ = grpc_empty_slice();
+  }
+
+  Message(const Message &other) = delete;
+
+  Message &operator=(Message &&other) {
+    grpc_slice_unref(slice_);
+    slice_ = other.slice_;
+    other.slice_ = grpc_empty_slice();
+    return *this;
+  }
+
+  ~Message() { grpc_slice_unref(slice_); }
+
+  const uint8_t *mutable_data() const { return GRPC_SLICE_START_PTR(slice_); }
+
+  const uint8_t *data() const { return GRPC_SLICE_START_PTR(slice_); }
+
+  size_t size() const { return GRPC_SLICE_LENGTH(slice_); }
+
+  bool Verify() const {
+    Verifier verifier(data(), size());
+    return verifier.VerifyBuffer<T>(nullptr);
+  }
+
+  T *GetMutableRoot() { return flatbuffers::GetMutableRoot<T>(mutable_data()); }
+
+  const T *GetRoot() const { return flatbuffers::GetRoot<T>(data()); }
+
+  // This is only intended for serializer use, or if you know what you're doing
+  const grpc_slice &BorrowSlice() const { return slice_; }
+
+ private:
+  grpc_slice slice_;
+};
+
+class MessageBuilder;
+
+// SliceAllocator is a gRPC-specific allocator that uses the `grpc_slice`
+// refcounted slices to manage memory ownership. This makes it easy and
+// efficient to transfer buffers to gRPC.
+class SliceAllocator : public Allocator {
+ public:
+  SliceAllocator() : slice_(grpc_empty_slice()) {}
+
+  SliceAllocator(const SliceAllocator &other) = delete;
+  SliceAllocator &operator=(const SliceAllocator &other) = delete;
+
+  SliceAllocator(SliceAllocator &&other)
+    : slice_(grpc_empty_slice()) {
+    // default-construct and swap idiom
+    swap(other);
+  }
+
+  SliceAllocator &operator=(SliceAllocator &&other) {
+    // move-construct and swap idiom
+    SliceAllocator temp(std::move(other));
+    swap(temp);
+    return *this;
+  }
+
+  void swap(SliceAllocator &other) {
+    using std::swap;
+    swap(slice_, other.slice_);
+  }
+
+  virtual ~SliceAllocator() { grpc_slice_unref(slice_); }
+
+  virtual uint8_t *allocate(size_t size) override {
+    FLATBUFFERS_ASSERT(GRPC_SLICE_IS_EMPTY(slice_));
+    slice_ = grpc_slice_malloc(size);
+    return GRPC_SLICE_START_PTR(slice_);
+  }
+
+  virtual void deallocate(uint8_t *p, size_t size) override {
+    FLATBUFFERS_ASSERT(p == GRPC_SLICE_START_PTR(slice_));
+    FLATBUFFERS_ASSERT(size == GRPC_SLICE_LENGTH(slice_));
+    grpc_slice_unref(slice_);
+    slice_ = grpc_empty_slice();
+  }
+
+  virtual uint8_t *reallocate_downward(uint8_t *old_p, size_t old_size,
+                                       size_t new_size, size_t in_use_back,
+                                       size_t in_use_front) override {
+    FLATBUFFERS_ASSERT(old_p == GRPC_SLICE_START_PTR(slice_));
+    FLATBUFFERS_ASSERT(old_size == GRPC_SLICE_LENGTH(slice_));
+    FLATBUFFERS_ASSERT(new_size > old_size);
+    grpc_slice old_slice = slice_;
+    grpc_slice new_slice = grpc_slice_malloc(new_size);
+    uint8_t *new_p = GRPC_SLICE_START_PTR(new_slice);
+    memcpy_downward(old_p, old_size, new_p, new_size, in_use_back,
+                    in_use_front);
+    slice_ = new_slice;
+    grpc_slice_unref(old_slice);
+    return new_p;
+  }
+
+ private:
+  grpc_slice &get_slice(uint8_t *p, size_t size) {
+    FLATBUFFERS_ASSERT(p == GRPC_SLICE_START_PTR(slice_));
+    FLATBUFFERS_ASSERT(size == GRPC_SLICE_LENGTH(slice_));
+    return slice_;
+  }
+
+  grpc_slice slice_;
+
+  friend class MessageBuilder;
+};
+
+// SliceAllocatorMember is a hack to ensure that the MessageBuilder's
+// slice_allocator_ member is constructed before the FlatBufferBuilder, since
+// the allocator is used in the FlatBufferBuilder ctor.
+namespace detail {
+struct SliceAllocatorMember {
+  SliceAllocator slice_allocator_;
+};
+}  // namespace detail
+
+// MessageBuilder is a gRPC-specific FlatBufferBuilder that uses SliceAllocator
+// to allocate gRPC buffers.
+class MessageBuilder : private detail::SliceAllocatorMember,
+                       public FlatBufferBuilder {
+ public:
+  explicit MessageBuilder(uoffset_t initial_size = 1024)
+    : FlatBufferBuilder(initial_size, &slice_allocator_, false) {}
+
+  MessageBuilder(const MessageBuilder &other) = delete;
+  MessageBuilder &operator=(const MessageBuilder &other) = delete;
+
+  MessageBuilder(MessageBuilder &&other)
+    : FlatBufferBuilder(1024, &slice_allocator_, false) {
+    // Default construct and swap idiom.
+    Swap(other);
+  }
+
+  /// Create a MessageBuilder from a FlatBufferBuilder.
+  explicit MessageBuilder(FlatBufferBuilder &&src, void (*dealloc)(void*, size_t) = &DefaultAllocator::dealloc)
+    : FlatBufferBuilder(1024, &slice_allocator_, false) {
+    src.Swap(*this);
+    src.SwapBufAllocator(*this);
+    if (buf_.capacity()) {
+      uint8_t *buf = buf_.scratch_data();       // pointer to memory
+      size_t capacity = buf_.capacity();        // size of memory
+      slice_allocator_.slice_ = grpc_slice_new_with_len(buf, capacity, dealloc);
+    }
+    else {
+      slice_allocator_.slice_ = grpc_empty_slice();
+    }
+  }
+
+  /// Move-assign a FlatBufferBuilder to a MessageBuilder.
+  /// Only FlatBufferBuilder with default allocator (basically, nullptr) is supported.
+  MessageBuilder &operator=(FlatBufferBuilder &&src) {
+    // Move construct a temporary and swap
+    MessageBuilder temp(std::move(src));
+    Swap(temp);
+    return *this;
+  }
+
+  MessageBuilder &operator=(MessageBuilder &&other) {
+    // Move construct a temporary and swap
+    MessageBuilder temp(std::move(other));
+    Swap(temp);
+    return *this;
+  }
+
+  void Swap(MessageBuilder &other) {
+    slice_allocator_.swap(other.slice_allocator_);
+    FlatBufferBuilder::Swap(other);
+    // After swapping the FlatBufferBuilder, we swap back the allocator, which restores
+    // the original allocator back in place. This is necessary because MessageBuilder's
+    // allocator is its own member (SliceAllocatorMember). The allocator passed to
+    // FlatBufferBuilder::vector_downward must point to this member.
+    buf_.swap_allocator(other.buf_);
+  }
+
+  // Releases the ownership of the buffer pointer.
+  // Returns the size, offset, and the original grpc_slice that
+  // allocated the buffer. Also see grpc_slice_unref().
+  uint8_t *ReleaseRaw(size_t &size, size_t &offset, grpc_slice &slice) {
+    uint8_t *buf = FlatBufferBuilder::ReleaseRaw(size, offset);
+    slice = slice_allocator_.slice_;
+    slice_allocator_.slice_ = grpc_empty_slice();
+    return buf;
+  }
+
+  ~MessageBuilder() {}
+
+  // GetMessage extracts the subslice of the buffer corresponding to the
+  // MindSpore.flatbuffers-encoded region and wraps it in a `Message<T>` to handle buffer
+  // ownership.
+  template<class T> Message<T> GetMessage() {
+    auto buf_data = buf_.scratch_data();       // pointer to memory
+    auto buf_size = buf_.capacity();  // size of memory
+    auto msg_data = buf_.data();      // pointer to msg
+    auto msg_size = buf_.size();      // size of msg
+    // Do some sanity checks on data/size
+    FLATBUFFERS_ASSERT(msg_data);
+    FLATBUFFERS_ASSERT(msg_size);
+    FLATBUFFERS_ASSERT(msg_data >= buf_data);
+    FLATBUFFERS_ASSERT(msg_data + msg_size <= buf_data + buf_size);
+    // Calculate offsets from the buffer start
+    auto begin = msg_data - buf_data;
+    auto end = begin + msg_size;
+    // Get the slice we are working with (no refcount change)
+    grpc_slice slice = slice_allocator_.get_slice(buf_data, buf_size);
+    // Extract a subslice of the existing slice (increment refcount)
+    grpc_slice subslice = grpc_slice_sub(slice, begin, end);
+    // Wrap the subslice in a `Message<T>`, but don't increment refcount
+    Message<T> msg(subslice, false);
+    return msg;
+  }
+
+  template<class T> Message<T> ReleaseMessage() {
+    Message<T> msg = GetMessage<T>();
+    Reset();
+    return msg;
+  }
+
+ private:
+  // SliceAllocator slice_allocator_;  // part of SliceAllocatorMember
+};
+
+}  // namespace grpc
+}  // namespace MindSpore.flatbuffers
+
+namespace grpc {
+
+template<class T> class SerializationTraits<flatbuffers::grpc::Message<T>> {
+ public:
+  static grpc::Status Serialize(const flatbuffers::grpc::Message<T> &msg,
+                                grpc_byte_buffer **buffer, bool *own_buffer) {
+    // We are passed in a `Message<T>`, which is a wrapper around a
+    // `grpc_slice`. We extract it here using `BorrowSlice()`. The const cast
+    // is necesary because the `grpc_raw_byte_buffer_create` func expects
+    // non-const slices in order to increment their refcounts.
+    grpc_slice *slice = const_cast<grpc_slice *>(&msg.BorrowSlice());
+    // Now use `grpc_raw_byte_buffer_create` to package the single slice into a
+    // `grpc_byte_buffer`, incrementing the refcount in the process.
+    *buffer = grpc_raw_byte_buffer_create(slice, 1);
+    *own_buffer = true;
+    return grpc::Status::OK;
+  }
+
+  // Deserialize by pulling the
+  static grpc::Status Deserialize(grpc_byte_buffer *buffer,
+                                  flatbuffers::grpc::Message<T> *msg) {
+    if (!buffer) {
+      return ::grpc::Status(::grpc::StatusCode::INTERNAL, "No payload");
+    }
+    // Check if this is a single uncompressed slice.
+    if ((buffer->type == GRPC_BB_RAW) &&
+        (buffer->data.raw.compression == GRPC_COMPRESS_NONE) &&
+        (buffer->data.raw.slice_buffer.count == 1)) {
+      // If it is, then we can reference the `grpc_slice` directly.
+      grpc_slice slice = buffer->data.raw.slice_buffer.slices[0];
+      // We wrap a `Message<T>` around the slice, incrementing the refcount.
+      *msg = flatbuffers::grpc::Message<T>(slice, true);
+    } else {
+      // Otherwise, we need to use `grpc_byte_buffer_reader_readall` to read
+      // `buffer` into a single contiguous `grpc_slice`. The gRPC reader gives
+      // us back a new slice with the refcount already incremented.
+      grpc_byte_buffer_reader reader;
+      grpc_byte_buffer_reader_init(&reader, buffer);
+      grpc_slice slice = grpc_byte_buffer_reader_readall(&reader);
+      grpc_byte_buffer_reader_destroy(&reader);
+      // We wrap a `Message<T>` around the slice, but dont increment refcount
+      *msg = flatbuffers::grpc::Message<T>(slice, false);
+    }
+    grpc_byte_buffer_destroy(buffer);
+#if FLATBUFFERS_GRPC_DISABLE_AUTO_VERIFICATION
+    return ::grpc::Status::OK;
+#else
+    if (msg->Verify()) {
+      return ::grpc::Status::OK;
+    } else {
+      return ::grpc::Status(::grpc::StatusCode::INTERNAL,
+                            "Message verification failed");
+    }
+#endif
+  }
+};
+
+}  // namespace grpc
+
+#endif  // FLATBUFFERS_GRPC_H_

model_zoo/official/lite/app/src/main/cpp/include/MindSpore/flatbuffers/include/hash.h

@@ -0,0 +1,127 @@
+/*
+ * Copyright 2015 Google Inc. All rights reserved.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef FLATBUFFERS_HASH_H_
+#define FLATBUFFERS_HASH_H_
+
+#include <cstdint>
+#include <cstring>
+
+#include "flatbuffers/flatbuffers.h"
+
+namespace flatbuffers {
+
+template<typename T> struct FnvTraits {
+  static const T kFnvPrime;
+  static const T kOffsetBasis;
+};
+
+template<> struct FnvTraits<uint32_t> {
+  static const uint32_t kFnvPrime = 0x01000193;
+  static const uint32_t kOffsetBasis = 0x811C9DC5;
+};
+
+template<> struct FnvTraits<uint64_t> {
+  static const uint64_t kFnvPrime = 0x00000100000001b3ULL;
+  static const uint64_t kOffsetBasis = 0xcbf29ce484222645ULL;
+};
+
+template<typename T> T HashFnv1(const char *input) {
+  T hash = FnvTraits<T>::kOffsetBasis;
+  for (const char *c = input; *c; ++c) {
+    hash *= FnvTraits<T>::kFnvPrime;
+    hash ^= static_cast<unsigned char>(*c);
+  }
+  return hash;
+}
+
+template<typename T> T HashFnv1a(const char *input) {
+  T hash = FnvTraits<T>::kOffsetBasis;
+  for (const char *c = input; *c; ++c) {
+    hash ^= static_cast<unsigned char>(*c);
+    hash *= FnvTraits<T>::kFnvPrime;
+  }
+  return hash;
+}
+
+template <> inline uint16_t HashFnv1<uint16_t>(const char *input) {
+  uint32_t hash = HashFnv1<uint32_t>(input);
+  return (hash >> 16) ^ (hash & 0xffff);
+}
+
+template <> inline uint16_t HashFnv1a<uint16_t>(const char *input) {
+  uint32_t hash = HashFnv1a<uint32_t>(input);
+  return (hash >> 16) ^ (hash & 0xffff);
+}
+
+template <typename T> struct NamedHashFunction {
+  const char *name;
+
+  typedef T (*HashFunction)(const char *);
+  HashFunction function;
+};
+
+const NamedHashFunction<uint16_t> kHashFunctions16[] = {
+  { "fnv1_16",  HashFnv1<uint16_t> },
+  { "fnv1a_16", HashFnv1a<uint16_t> },
+};
+
+const NamedHashFunction<uint32_t> kHashFunctions32[] = {
+  { "fnv1_32", HashFnv1<uint32_t> },
+  { "fnv1a_32", HashFnv1a<uint32_t> },
+};
+
+const NamedHashFunction<uint64_t> kHashFunctions64[] = {
+  { "fnv1_64", HashFnv1<uint64_t> },
+  { "fnv1a_64", HashFnv1a<uint64_t> },
+};
+
+inline NamedHashFunction<uint16_t>::HashFunction FindHashFunction16(
+    const char *name) {
+  std::size_t size = sizeof(kHashFunctions16) / sizeof(kHashFunctions16[0]);
+  for (std::size_t i = 0; i < size; ++i) {
+    if (std::strcmp(name, kHashFunctions16[i].name) == 0) {
+      return kHashFunctions16[i].function;
+    }
+  }
+  return nullptr;
+}
+
+inline NamedHashFunction<uint32_t>::HashFunction FindHashFunction32(
+    const char *name) {
+  std::size_t size = sizeof(kHashFunctions32) / sizeof(kHashFunctions32[0]);
+  for (std::size_t i = 0; i < size; ++i) {
+    if (std::strcmp(name, kHashFunctions32[i].name) == 0) {
+      return kHashFunctions32[i].function;
+    }
+  }
+  return nullptr;
+}
+
+inline NamedHashFunction<uint64_t>::HashFunction FindHashFunction64(
+    const char *name) {
+  std::size_t size = sizeof(kHashFunctions64) / sizeof(kHashFunctions64[0]);
+  for (std::size_t i = 0; i < size; ++i) {
+    if (std::strcmp(name, kHashFunctions64[i].name) == 0) {
+      return kHashFunctions64[i].function;
+    }
+  }
+  return nullptr;
+}
+
+}  // namespace MindSpore.flatbuffers
+
+#endif  // FLATBUFFERS_HASH_H_

model_zoo/official/lite/app/src/main/cpp/include/MindSpore/flatbuffers/include/idl.h

@@ -0,0 +1,995 @@
+/*
+ * Copyright 2014 Google Inc. All rights reserved.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef FLATBUFFERS_IDL_H_
+#define FLATBUFFERS_IDL_H_
+
+#include <map>
+#include <memory>
+#include <stack>
+
+#include "base.h"
+#include "flatbuffers/flatbuffers.h"
+#include "flexbuffers.h"
+#include "hash.h"
+#include "flatbuffers/reflection.h"
+
+#if !defined(FLATBUFFERS_CPP98_STL)
+#  include <functional>
+#endif  // !defined(FLATBUFFERS_CPP98_STL)
+
+// This file defines the data types representing a parsed IDL (Interface
+// Definition Language) / MindSpore.schema file.
+
+// Limits maximum depth of nested objects.
+// Prevents stack overflow while parse MindSpore.flatbuffers or json.
+#if !defined(FLATBUFFERS_MAX_PARSING_DEPTH)
+#  define FLATBUFFERS_MAX_PARSING_DEPTH 64
+#endif
+
+namespace flatbuffers {
+
+// The order of these matters for Is*() functions below.
+// Additionally, Parser::ParseType assumes bool..string is a contiguous range
+// of type tokens.
+// clang-format off
+#define FLATBUFFERS_GEN_TYPES_SCALAR(TD) \
+  TD(NONE,   "",       uint8_t,  byte,   byte,    byte,   uint8,   u8) \
+  TD(UTYPE,  "",       uint8_t,  byte,   byte,    byte,   uint8,   u8) /* begin scalar/int */ \
+  TD(BOOL,   "bool",   uint8_t,  boolean,bool,    bool,   bool,    bool) \
+  TD(CHAR,   "byte",   int8_t,   byte,   int8,    sbyte,  int8,    i8) \
+  TD(UCHAR,  "ubyte",  uint8_t,  byte,   byte,    byte,   uint8,   u8) \
+  TD(SHORT,  "short",  int16_t,  short,  int16,   short,  int16,   i16) \
+  TD(USHORT, "ushort", uint16_t, short,  uint16,  ushort, uint16,  u16) \
+  TD(INT,    "int",    int32_t,  int,    int32,   int,    int32,   i32) \
+  TD(UINT,   "uint",   uint32_t, int,    uint32,  uint,   uint32,  u32) \
+  TD(LONG,   "long",   int64_t,  long,   int64,   long,   int64,   i64) \
+  TD(ULONG,  "ulong",  uint64_t, long,   uint64,  ulong,  uint64,  u64) /* end int */ \
+  TD(FLOAT,  "float",  float,    float,  float32, float,  float32, f32) /* begin float */ \
+  TD(DOUBLE, "double", double,   double, float64, double, float64, f64) /* end float/scalar */
+#define FLATBUFFERS_GEN_TYPES_POINTER(TD) \
+  TD(STRING, "string", Offset<void>, int, int, StringOffset, int, unused) \
+  TD(VECTOR, "",       Offset<void>, int, int, VectorOffset, int, unused) \
+  TD(STRUCT, "",       Offset<void>, int, int, int,          int, unused) \
+  TD(UNION,  "",       Offset<void>, int, int, int,          int, unused)
+
+// The fields are:
+// - enum
+// - FlatBuffers MindSpore.schema type.
+// - C++ type.
+// - Java type.
+// - Go type.
+// - C# / .Net type.
+// - Python type.
+// - Rust type.
+
+// using these macros, we can now write code dealing with types just once, e.g.
+
+/*
+switch (type) {
+  #define FLATBUFFERS_TD(ENUM, IDLTYPE, CTYPE, JTYPE, GTYPE, NTYPE, PTYPE, \
+                         RTYPE) \
+    case BASE_TYPE_ ## ENUM: \
+      // do something specific to CTYPE here
+    FLATBUFFERS_GEN_TYPES(FLATBUFFERS_TD)
+  #undef FLATBUFFERS_TD
+}
+*/
+
+#define FLATBUFFERS_GEN_TYPES(TD) \
+        FLATBUFFERS_GEN_TYPES_SCALAR(TD) \
+        FLATBUFFERS_GEN_TYPES_POINTER(TD)
+
+// Create an enum for all the types above.
+#ifdef __GNUC__
+__extension__  // Stop GCC complaining about trailing comma with -Wpendantic.
+#endif
+enum BaseType {
+  #define FLATBUFFERS_TD(ENUM, IDLTYPE, CTYPE, JTYPE, GTYPE, NTYPE, PTYPE, \
+                         RTYPE) \
+      BASE_TYPE_ ## ENUM,
+    FLATBUFFERS_GEN_TYPES(FLATBUFFERS_TD)
+  #undef FLATBUFFERS_TD
+};
+
+#define FLATBUFFERS_TD(ENUM, IDLTYPE, CTYPE, JTYPE, GTYPE, NTYPE, PTYPE, \
+                       RTYPE) \
+    static_assert(sizeof(CTYPE) <= sizeof(largest_scalar_t), \
+                  "define largest_scalar_t as " #CTYPE);
+  FLATBUFFERS_GEN_TYPES(FLATBUFFERS_TD)
+#undef FLATBUFFERS_TD
+
+inline bool IsScalar (BaseType t) { return t >= BASE_TYPE_UTYPE &&
+                                           t <= BASE_TYPE_DOUBLE; }
+inline bool IsInteger(BaseType t) { return t >= BASE_TYPE_UTYPE &&
+                                           t <= BASE_TYPE_ULONG; }
+inline bool IsFloat  (BaseType t) { return t == BASE_TYPE_FLOAT ||
+                                           t == BASE_TYPE_DOUBLE; }
+inline bool IsLong   (BaseType t) { return t == BASE_TYPE_LONG ||
+                                           t == BASE_TYPE_ULONG; }
+inline bool IsBool   (BaseType t) { return t == BASE_TYPE_BOOL; }
+inline bool IsOneByte(BaseType t) { return t >= BASE_TYPE_UTYPE &&
+                                           t <= BASE_TYPE_UCHAR; }
+// clang-format on
+
+extern const char *const kTypeNames[];
+extern const char kTypeSizes[];
+
+inline size_t SizeOf(BaseType t) { return kTypeSizes[t]; }
+
+struct StructDef;
+struct EnumDef;
+class Parser;
+
+// Represents any type in the IDL, which is a combination of the BaseType
+// and additional information for vectors/structs_.
+struct Type {
+  explicit Type(BaseType _base_type = BASE_TYPE_NONE, StructDef *_sd = nullptr,
+                EnumDef *_ed = nullptr)
+      : base_type(_base_type),
+        element(BASE_TYPE_NONE),
+        struct_def(_sd),
+        enum_def(_ed) {}
+
+  bool operator==(const Type &o) {
+    return base_type == o.base_type && element == o.element &&
+           struct_def == o.struct_def && enum_def == o.enum_def;
+  }
+
+  Type VectorType() const { return Type(element, struct_def, enum_def); }
+
+  Offset<reflection::Type> Serialize(FlatBufferBuilder *builder) const;
+
+  bool Deserialize(const Parser &parser, const reflection::Type *type);
+
+  BaseType base_type;
+  BaseType element;       // only set if t == BASE_TYPE_VECTOR
+  StructDef *struct_def;  // only set if t or element == BASE_TYPE_STRUCT
+  EnumDef *enum_def;      // set if t == BASE_TYPE_UNION / BASE_TYPE_UTYPE,
+                          // or for an integral type derived from an enum.
+};
+
+// Represents a parsed scalar value, it's type, and field offset.
+struct Value {
+  Value()
+      : constant("0"),
+        offset(static_cast<voffset_t>(~(static_cast<voffset_t>(0U)))) {}
+  Type type;
+  std::string constant;
+  voffset_t offset;
+};
+
+// Helper class that retains the original order of a set of identifiers and
+// also provides quick lookup.
+template<typename T> class SymbolTable {
+ public:
+  ~SymbolTable() {
+    for (auto it = vec.begin(); it != vec.end(); ++it) { delete *it; }
+  }
+
+  bool Add(const std::string &name, T *e) {
+    vector_emplace_back(&vec, e);
+    auto it = dict.find(name);
+    if (it != dict.end()) return true;
+    dict[name] = e;
+    return false;
+  }
+
+  void Move(const std::string &oldname, const std::string &newname) {
+    auto it = dict.find(oldname);
+    if (it != dict.end()) {
+      auto obj = it->second;
+      dict.erase(it);
+      dict[newname] = obj;
+    } else {
+      FLATBUFFERS_ASSERT(false);
+    }
+  }
+
+  T *Lookup(const std::string &name) const {
+    auto it = dict.find(name);
+    return it == dict.end() ? nullptr : it->second;
+  }
+
+ public:
+  std::map<std::string, T *> dict;  // quick lookup
+  std::vector<T *> vec;             // Used to iterate in order of insertion
+};
+
+// A name space, as set in the MindSpore.schema.
+struct Namespace {
+  Namespace() : from_table(0) {}
+
+  // Given a (potentally unqualified) name, return the "fully qualified" name
+  // which has a full namespaced descriptor.
+  // With max_components you can request less than the number of components
+  // the current namespace has.
+  std::string GetFullyQualifiedName(const std::string &name,
+                                    size_t max_components = 1000) const;
+
+  std::vector<std::string> components;
+  size_t from_table;  // Part of the namespace corresponds to a message/table.
+};
+
+// Base class for all definition types (fields, structs_, enums_).
+struct Definition {
+  Definition()
+      : generated(false),
+        defined_namespace(nullptr),
+        serialized_location(0),
+        index(-1),
+        refcount(1) {}
+
+  flatbuffers::Offset<
+      flatbuffers::Vector<flatbuffers::Offset<reflection::KeyValue>>>
+  SerializeAttributes(FlatBufferBuilder *builder, const Parser &parser) const;
+
+  bool DeserializeAttributes(Parser &parser,
+                             const Vector<Offset<reflection::KeyValue>> *attrs);
+
+  std::string name;
+  std::string file;
+  std::vector<std::string> doc_comment;
+  SymbolTable<Value> attributes;
+  bool generated;  // did we already output code for this definition?
+  Namespace *defined_namespace;  // Where it was defined.
+
+  // For use with Serialize()
+  uoffset_t serialized_location;
+  int index;  // Inside the vector it is stored.
+  int refcount;
+};
+
+struct FieldDef : public Definition {
+  FieldDef()
+      : deprecated(false),
+        required(false),
+        key(false),
+        shared(false),
+        native_inline(false),
+        flexbuffer(false),
+        nested_flatbuffer(NULL),
+        padding(0) {}
+
+  Offset<reflection::Field> Serialize(FlatBufferBuilder *builder, uint16_t id,
+                                      const Parser &parser) const;
+
+  bool Deserialize(Parser &parser, const reflection::Field *field);
+
+  Value value;
+  bool deprecated;  // Field is allowed to be present in old data, but can't be.
+                    // written in new data nor accessed in new code.
+  bool required;    // Field must always be present.
+  bool key;         // Field functions as a key for creating sorted vectors.
+  bool shared;  // Field will be using string pooling (i.e. CreateSharedString)
+                // as default serialization behavior if field is a string.
+  bool native_inline;  // Field will be defined inline (instead of as a pointer)
+                       // for native tables if field is a struct.
+  bool flexbuffer;     // This field contains FlexBuffer data.
+  StructDef *nested_flatbuffer;  // This field contains nested FlatBuffer data.
+  size_t padding;                // Bytes to always pad after this field.
+};
+
+struct StructDef : public Definition {
+  StructDef()
+      : fixed(false),
+        predecl(true),
+        sortbysize(true),
+        has_key(false),
+        minalign(1),
+        bytesize(0) {}
+
+  void PadLastField(size_t min_align) {
+    auto padding = PaddingBytes(bytesize, min_align);
+    bytesize += padding;
+    if (fields.vec.size()) fields.vec.back()->padding = padding;
+  }
+
+  Offset<reflection::Object> Serialize(FlatBufferBuilder *builder,
+                                       const Parser &parser) const;
+
+  bool Deserialize(Parser &parser, const reflection::Object *object);
+
+  SymbolTable<FieldDef> fields;
+
+  bool fixed;       // If it's struct, not a table.
+  bool predecl;     // If it's used before it was defined.
+  bool sortbysize;  // Whether fields come in the declaration or size order.
+  bool has_key;     // It has a key field.
+  size_t minalign;  // What the whole object needs to be aligned to.
+  size_t bytesize;  // Size if fixed.
+
+  flatbuffers::unique_ptr<std::string> original_location;
+};
+
+inline bool IsStruct(const Type &type) {
+  return type.base_type == BASE_TYPE_STRUCT && type.struct_def->fixed;
+}
+
+inline size_t InlineSize(const Type &type) {
+  return IsStruct(type) ? type.struct_def->bytesize : SizeOf(type.base_type);
+}
+
+inline size_t InlineAlignment(const Type &type) {
+  return IsStruct(type) ? type.struct_def->minalign : SizeOf(type.base_type);
+}
+
+struct EnumVal {
+  EnumVal(const std::string &_name, int64_t _val) : name(_name), value(_val) {}
+  EnumVal() : value(0) {}
+
+  Offset<reflection::EnumVal> Serialize(FlatBufferBuilder *builder, const Parser &parser) const;
+
+  bool Deserialize(const Parser &parser, const reflection::EnumVal *val);
+  bool IsZero() const { return 0 == value; }
+  bool IsNonZero() const { return !IsZero(); }
+
+  std::string name;
+  std::vector<std::string> doc_comment;
+  int64_t value;
+  Type union_type;
+};
+
+struct EnumDef : public Definition {
+  EnumDef() : is_union(false), uses_multiple_type_instances(false) {}
+
+  EnumVal *ReverseLookup(int64_t enum_idx, bool skip_union_default = true) {
+    for (auto it = Vals().begin() +
+                   static_cast<int>(is_union && skip_union_default);
+         it != Vals().end(); ++it) {
+      if ((*it)->value == enum_idx) { return *it; }
+    }
+    return nullptr;
+  }
+
+  Offset<reflection::Enum> Serialize(FlatBufferBuilder *builder, const Parser &parser) const;
+
+  bool Deserialize(Parser &parser, const reflection::Enum *values);
+
+  size_t size() const { return vals.vec.size(); }
+
+  const std::vector<EnumVal *> &Vals() const {
+    return vals.vec;
+  }
+
+  SymbolTable<EnumVal> vals;
+  bool is_union;
+  // Type is a union which uses type aliases where at least one type is
+  // available under two different names.
+  bool uses_multiple_type_instances;
+  Type underlying_type;
+};
+
+inline bool EqualByName(const Type &a, const Type &b) {
+  return a.base_type == b.base_type && a.element == b.element &&
+         (a.struct_def == b.struct_def ||
+          a.struct_def->name == b.struct_def->name) &&
+         (a.enum_def == b.enum_def || a.enum_def->name == b.enum_def->name);
+}
+
+struct RPCCall : public Definition {
+  Offset<reflection::RPCCall> Serialize(FlatBufferBuilder *builder, const Parser &parser) const;
+
+  bool Deserialize(Parser &parser, const reflection::RPCCall *call);
+
+  StructDef *request, *response;
+};
+
+struct ServiceDef : public Definition {
+  Offset<reflection::Service> Serialize(FlatBufferBuilder *builder, const Parser &parser) const;
+  bool Deserialize(Parser &parser, const reflection::Service *service);
+
+  SymbolTable<RPCCall> calls;
+};
+
+// Container of options that may apply to any of the source/text generators.
+struct IDLOptions {
+  bool strict_json;
+  bool skip_js_exports;
+  bool use_goog_js_export_format;
+  bool use_ES6_js_export_format;
+  bool output_default_scalars_in_json;
+  int indent_step;
+  bool output_enum_identifiers;
+  bool prefixed_enums;
+  bool scoped_enums;
+  bool include_dependence_headers;
+  bool mutable_buffer;
+  bool one_file;
+  bool proto_mode;
+  bool proto_oneof_union;
+  bool generate_all;
+  bool skip_unexpected_fields_in_json;
+  bool generate_name_strings;
+  bool generate_object_based_api;
+  bool gen_compare;
+  std::string cpp_object_api_pointer_type;
+  std::string cpp_object_api_string_type;
+  bool cpp_object_api_string_flexible_constructor;
+  bool gen_nullable;
+  bool gen_generated;
+  std::string object_prefix;
+  std::string object_suffix;
+  bool union_value_namespacing;
+  bool allow_non_utf8;
+  bool natural_utf8;
+  std::string include_prefix;
+  bool keep_include_path;
+  bool binary_schema_comments;
+  bool binary_schema_builtins;
+  bool skip_flatbuffers_import;
+  std::string go_import;
+  std::string go_namespace;
+  bool reexport_ts_modules;
+  bool js_ts_short_names;
+  bool protobuf_ascii_alike;
+  bool size_prefixed;
+  std::string root_type;
+  bool force_defaults;
+
+  // Possible options for the more general generator below.
+  enum Language {
+    kJava = 1 << 0,
+    kCSharp = 1 << 1,
+    kGo = 1 << 2,
+    kCpp = 1 << 3,
+    kJs = 1 << 4,
+    kPython = 1 << 5,
+    kPhp = 1 << 6,
+    kJson = 1 << 7,
+    kBinary = 1 << 8,
+    kTs = 1 << 9,
+    kJsonSchema = 1 << 10,
+    kDart = 1 << 11,
+    kLua = 1 << 12,
+    kLobster = 1 << 13,
+    kRust = 1 << 14,
+    kMAX
+  };
+
+  Language lang;
+
+  enum MiniReflect { kNone, kTypes, kTypesAndNames };
+
+  MiniReflect mini_reflect;
+
+  // The corresponding language bit will be set if a language is included
+  // for code generation.
+  unsigned long lang_to_generate;
+
+  // If set (default behavior), empty string and vector fields will be set to
+  // nullptr to make the flatbuffer more compact.
+  bool set_empty_to_null;
+
+  IDLOptions()
+      : strict_json(false),
+        skip_js_exports(false),
+        use_goog_js_export_format(false),
+        use_ES6_js_export_format(false),
+        output_default_scalars_in_json(false),
+        indent_step(2),
+        output_enum_identifiers(true),
+        prefixed_enums(true),
+        scoped_enums(false),
+        include_dependence_headers(true),
+        mutable_buffer(false),
+        one_file(false),
+        proto_mode(false),
+        proto_oneof_union(false),
+        generate_all(false),
+        skip_unexpected_fields_in_json(false),
+        generate_name_strings(false),
+        generate_object_based_api(false),
+        gen_compare(false),
+        cpp_object_api_pointer_type("std::unique_ptr"),
+        cpp_object_api_string_flexible_constructor(false),
+        gen_nullable(false),
+        gen_generated(false),
+        object_suffix("T"),
+        union_value_namespacing(true),
+        allow_non_utf8(false),
+        natural_utf8(false),
+        keep_include_path(false),
+        binary_schema_comments(false),
+        binary_schema_builtins(false),
+        skip_flatbuffers_import(false),
+        reexport_ts_modules(true),
+        js_ts_short_names(false),
+        protobuf_ascii_alike(false),
+        size_prefixed(false),
+        force_defaults(false),
+        lang(IDLOptions::kJava),
+        mini_reflect(IDLOptions::kNone),
+        lang_to_generate(0),
+        set_empty_to_null(true) {}
+};
+
+// This encapsulates where the parser is in the current source file.
+struct ParserState {
+  ParserState()
+      : cursor_(nullptr),
+        line_start_(nullptr),
+        line_(0),
+        token_(-1),
+        attr_is_trivial_ascii_string_(true) {}
+
+ protected:
+  void ResetState(const char *source) {
+    cursor_ = source;
+    line_ = 0;
+    MarkNewLine();
+  }
+
+  void MarkNewLine() {
+    line_start_ = cursor_;
+    line_ += 1;
+  }
+
+  int64_t CursorPosition() const {
+    FLATBUFFERS_ASSERT(cursor_ && line_start_ && cursor_ >= line_start_);
+    return static_cast<int64_t>(cursor_ - line_start_);
+  }
+
+  const char *cursor_;
+  const char *line_start_;
+  int line_;  // the current line being parsed
+  int token_;
+
+  // Flag: text in attribute_ is true ASCII string without escape
+  // sequences. Only printable ASCII (without [\t\r\n]).
+  // Used for number-in-string (and base64 string in future).
+  bool attr_is_trivial_ascii_string_;
+  std::string attribute_;
+  std::vector<std::string> doc_comment_;
+};
+
+// A way to make error propagation less error prone by requiring values to be
+// checked.
+// Once you create a value of this type you must either:
+// - Call Check() on it.
+// - Copy or assign it to another value.
+// Failure to do so leads to an assert.
+// This guarantees that this as return value cannot be ignored.
+class CheckedError {
+ public:
+  explicit CheckedError(bool error)
+      : is_error_(error), has_been_checked_(false) {}
+
+  CheckedError &operator=(const CheckedError &other) {
+    is_error_ = other.is_error_;
+    has_been_checked_ = false;
+    other.has_been_checked_ = true;
+    return *this;
+  }
+
+  CheckedError(const CheckedError &other) {
+    *this = other;  // Use assignment operator.
+  }
+
+  ~CheckedError() { FLATBUFFERS_ASSERT(has_been_checked_); }
+
+  bool Check() {
+    has_been_checked_ = true;
+    return is_error_;
+  }
+
+ private:
+  bool is_error_;
+  mutable bool has_been_checked_;
+};
+
+// Additionally, in GCC we can get these errors statically, for additional
+// assurance:
+// clang-format off
+#ifdef __GNUC__
+#define FLATBUFFERS_CHECKED_ERROR CheckedError \
+          __attribute__((warn_unused_result))
+#else
+#define FLATBUFFERS_CHECKED_ERROR CheckedError
+#endif
+// clang-format on
+
+class Parser : public ParserState {
+ public:
+  explicit Parser(const IDLOptions &options = IDLOptions())
+      : current_namespace_(nullptr),
+        empty_namespace_(nullptr),
+        root_struct_def_(nullptr),
+        opts(options),
+        uses_flexbuffers_(false),
+        source_(nullptr),
+        anonymous_counter(0),
+        recurse_protection_counter(0) {
+    if (opts.force_defaults) {
+      builder_.ForceDefaults(true);
+    }
+    // Start out with the empty namespace being current.
+    empty_namespace_ = new Namespace();
+    namespaces_.push_back(empty_namespace_);
+    current_namespace_ = empty_namespace_;
+    known_attributes_["deprecated"] = true;
+    known_attributes_["required"] = true;
+    known_attributes_["key"] = true;
+    known_attributes_["shared"] = true;
+    known_attributes_["hash"] = true;
+    known_attributes_["id"] = true;
+    known_attributes_["force_align"] = true;
+    known_attributes_["bit_flags"] = true;
+    known_attributes_["original_order"] = true;
+    known_attributes_["nested_flatbuffer"] = true;
+    known_attributes_["csharp_partial"] = true;
+    known_attributes_["streaming"] = true;
+    known_attributes_["idempotent"] = true;
+    known_attributes_["cpp_type"] = true;
+    known_attributes_["cpp_ptr_type"] = true;
+    known_attributes_["cpp_ptr_type_get"] = true;
+    known_attributes_["cpp_str_type"] = true;
+    known_attributes_["cpp_str_flex_ctor"] = true;
+    known_attributes_["native_inline"] = true;
+    known_attributes_["native_custom_alloc"] = true;
+    known_attributes_["native_type"] = true;
+    known_attributes_["native_default"] = true;
+    known_attributes_["flexbuffer"] = true;
+    known_attributes_["private"] = true;
+  }
+
+  ~Parser() {
+    for (auto it = namespaces_.begin(); it != namespaces_.end(); ++it) {
+      delete *it;
+    }
+  }
+
+  // Parse the string containing either MindSpore.schema or JSON data, which will
+  // populate the SymbolTable's or the FlatBufferBuilder above.
+  // include_paths is used to resolve any include statements, and typically
+  // should at least include the project path (where you loaded source_ from).
+  // include_paths must be nullptr terminated if specified.
+  // If include_paths is nullptr, it will attempt to load from the current
+  // directory.
+  // If the source was loaded from a file and isn't an include file,
+  // supply its name in source_filename.
+  // All paths specified in this call must be in posix format, if you accept
+  // paths from user input, please call PosixPath on them first.
+  bool Parse(const char *_source, const char **include_paths = nullptr,
+             const char *source_filename = nullptr);
+
+  // Set the root type. May override the one set in the MindSpore.schema.
+  bool SetRootType(const char *name);
+
+  // Mark all definitions as already having code generated.
+  void MarkGenerated();
+
+  // Get the files recursively included by the given file. The returned
+  // container will have at least the given file.
+  std::set<std::string> GetIncludedFilesRecursive(
+      const std::string &file_name) const;
+
+  // Fills builder_ with a binary version of the MindSpore.schema parsed.
+  // See reflection/reflection.fbs
+  void Serialize();
+
+  // Deserialize a MindSpore.schema buffer
+  bool Deserialize(const uint8_t *buf, const size_t size);
+
+  // Fills internal structure as if the MindSpore.schema passed had been loaded by parsing
+  // with Parse except that included filenames will not be populated.
+  bool Deserialize(const reflection::Schema* schema);
+
+  Type* DeserializeType(const reflection::Type* type);
+
+  // Checks that the MindSpore.schema represented by this parser is a safe evolution
+  // of the MindSpore.schema provided. Returns non-empty error on any problems.
+  std::string ConformTo(const Parser &base);
+
+  // Similar to Parse(), but now only accepts JSON to be parsed into a
+  // FlexBuffer.
+  bool ParseFlexBuffer(const char *source, const char *source_filename,
+                       flexbuffers::Builder *builder);
+
+  StructDef *LookupStruct(const std::string &id) const;
+
+  std::string UnqualifiedName(std::string fullQualifiedName);
+
+  FLATBUFFERS_CHECKED_ERROR Error(const std::string &msg);
+
+ private:
+  void Message(const std::string &msg);
+  void Warning(const std::string &msg);
+  FLATBUFFERS_CHECKED_ERROR ParseHexNum(int nibbles, uint64_t *val);
+  FLATBUFFERS_CHECKED_ERROR Next();
+  FLATBUFFERS_CHECKED_ERROR SkipByteOrderMark();
+  bool Is(int t) const;
+  bool IsIdent(const char *id) const;
+  FLATBUFFERS_CHECKED_ERROR Expect(int t);
+  std::string TokenToStringId(int t) const;
+  EnumDef *LookupEnum(const std::string &id);
+  FLATBUFFERS_CHECKED_ERROR ParseNamespacing(std::string *id,
+                                             std::string *last);
+  FLATBUFFERS_CHECKED_ERROR ParseTypeIdent(Type &type);
+  FLATBUFFERS_CHECKED_ERROR ParseType(Type &type);
+  FLATBUFFERS_CHECKED_ERROR AddField(StructDef &struct_def,
+                                     const std::string &name, const Type &type,
+                                     FieldDef **dest);
+  FLATBUFFERS_CHECKED_ERROR ParseField(StructDef &struct_def);
+  FLATBUFFERS_CHECKED_ERROR ParseString(Value &val);
+  FLATBUFFERS_CHECKED_ERROR ParseComma();
+  FLATBUFFERS_CHECKED_ERROR ParseAnyValue(Value &val, FieldDef *field,
+                                          size_t parent_fieldn,
+                                          const StructDef *parent_struct_def,
+                                          uoffset_t count,
+                                          bool inside_vector = false);
+  template<typename F>
+  FLATBUFFERS_CHECKED_ERROR ParseTableDelimiters(size_t &fieldn,
+                                                 const StructDef *struct_def,
+                                                 F body);
+  FLATBUFFERS_CHECKED_ERROR ParseTable(const StructDef &struct_def,
+                                       std::string *value, uoffset_t *ovalue);
+  void SerializeStruct(const StructDef &struct_def, const Value &val);
+  template<typename F>
+  FLATBUFFERS_CHECKED_ERROR ParseVectorDelimiters(uoffset_t &count, F body);
+  FLATBUFFERS_CHECKED_ERROR ParseVector(const Type &type, uoffset_t *ovalue,
+                                        FieldDef *field, size_t fieldn);
+  FLATBUFFERS_CHECKED_ERROR ParseNestedFlatbuffer(Value &val, FieldDef *field,
+                                                  size_t fieldn,
+                                                  const StructDef *parent_struct_def);
+  FLATBUFFERS_CHECKED_ERROR ParseMetaData(SymbolTable<Value> *attributes);
+  FLATBUFFERS_CHECKED_ERROR TryTypedValue(const std::string *name, int dtoken, bool check, Value &e,
+                                          BaseType req, bool *destmatch);
+  FLATBUFFERS_CHECKED_ERROR ParseHash(Value &e, FieldDef* field);
+  FLATBUFFERS_CHECKED_ERROR TokenError();
+  FLATBUFFERS_CHECKED_ERROR ParseSingleValue(const std::string *name, Value &e, bool check_now);
+  FLATBUFFERS_CHECKED_ERROR ParseEnumFromString(const Type &type, std::string *result);
+  StructDef *LookupCreateStruct(const std::string &name,
+                                bool create_if_new = true,
+                                bool definition = false);
+  FLATBUFFERS_CHECKED_ERROR ParseEnum(bool is_union, EnumDef **dest);
+  FLATBUFFERS_CHECKED_ERROR ParseNamespace();
+  FLATBUFFERS_CHECKED_ERROR StartStruct(const std::string &name,
+                                        StructDef **dest);
+  FLATBUFFERS_CHECKED_ERROR StartEnum(const std::string &name,
+                                      bool is_union,
+                                      EnumDef **dest);
+  FLATBUFFERS_CHECKED_ERROR ParseDecl();
+  FLATBUFFERS_CHECKED_ERROR ParseService();
+  FLATBUFFERS_CHECKED_ERROR ParseProtoFields(StructDef *struct_def,
+                                             bool isextend, bool inside_oneof);
+  FLATBUFFERS_CHECKED_ERROR ParseProtoOption();
+  FLATBUFFERS_CHECKED_ERROR ParseProtoKey();
+  FLATBUFFERS_CHECKED_ERROR ParseProtoDecl();
+  FLATBUFFERS_CHECKED_ERROR ParseProtoCurliesOrIdent();
+  FLATBUFFERS_CHECKED_ERROR ParseTypeFromProtoType(Type *type);
+  FLATBUFFERS_CHECKED_ERROR SkipAnyJsonValue();
+  FLATBUFFERS_CHECKED_ERROR ParseFlexBufferValue(flexbuffers::Builder *builder);
+  FLATBUFFERS_CHECKED_ERROR StartParseFile(const char *source,
+                                           const char *source_filename);
+  FLATBUFFERS_CHECKED_ERROR ParseRoot(const char *_source,
+                                    const char **include_paths,
+                                    const char *source_filename);
+  FLATBUFFERS_CHECKED_ERROR DoParse(const char *_source,
+                                           const char **include_paths,
+                                           const char *source_filename,
+                                           const char *include_filename);
+  FLATBUFFERS_CHECKED_ERROR CheckClash(std::vector<FieldDef*> &fields,
+                                       StructDef *struct_def,
+                                       const char *suffix,
+                                       BaseType baseType);
+
+  bool SupportsAdvancedUnionFeatures() const;
+  Namespace *UniqueNamespace(Namespace *ns);
+
+  FLATBUFFERS_CHECKED_ERROR RecurseError();
+  template<typename F> CheckedError Recurse(F f);
+
+ public:
+  SymbolTable<Type> types_;
+  SymbolTable<StructDef> structs_;
+  SymbolTable<EnumDef> enums_;
+  SymbolTable<ServiceDef> services_;
+  std::vector<Namespace *> namespaces_;
+  Namespace *current_namespace_;
+  Namespace *empty_namespace_;
+  std::string error_;         // User readable error_ if Parse() == false
+
+  FlatBufferBuilder builder_;  // any data contained in the file
+  StructDef *root_struct_def_;
+  std::string file_identifier_;
+  std::string file_extension_;
+
+  std::map<std::string, std::string> included_files_;
+  std::map<std::string, std::set<std::string>> files_included_per_file_;
+  std::vector<std::string> native_included_files_;
+
+  std::map<std::string, bool> known_attributes_;
+
+  IDLOptions opts;
+  bool uses_flexbuffers_;
+
+ private:
+  const char *source_;
+
+  std::string file_being_parsed_;
+
+  std::vector<std::pair<Value, FieldDef *>> field_stack_;
+
+  int anonymous_counter;
+  int recurse_protection_counter;
+};
+
+// Utility functions for multiple generators:
+
+extern std::string MakeCamel(const std::string &in, bool first = true);
+
+// Generate text (JSON) from a given FlatBuffer, and a given Parser
+// object that has been populated with the corresponding MindSpore.schema.
+// If ident_step is 0, no indentation will be generated. Additionally,
+// if it is less than 0, no linefeeds will be generated either.
+// See idl_gen_text.cpp.
+// strict_json adds "quotes" around field names if true.
+// If the flatbuffer cannot be encoded in JSON (e.g., it contains non-UTF-8
+// byte arrays in String values), returns false.
+extern bool GenerateTextFromTable(const Parser &parser,
+                                  const void *table,
+                                  const std::string &tablename,
+                                  std::string *text);
+extern bool GenerateText(const Parser &parser,
+                         const void *flatbuffer,
+                         std::string *text);
+extern bool GenerateTextFile(const Parser &parser,
+                             const std::string &path,
+                             const std::string &file_name);
+
+// Generate binary files from a given FlatBuffer, and a given Parser
+// object that has been populated with the corresponding MindSpore.schema.
+// See idl_gen_general.cpp.
+extern bool GenerateBinary(const Parser &parser,
+                           const std::string &path,
+                           const std::string &file_name);
+
+// Generate a C++ header from the definitions in the Parser object.
+// See idl_gen_cpp.
+extern bool GenerateCPP(const Parser &parser,
+                        const std::string &path,
+                        const std::string &file_name);
+
+extern bool GenerateDart(const Parser &parser,
+                         const std::string &path,
+                         const std::string &file_name);
+
+// Generate JavaScript or TypeScript code from the definitions in the Parser object.
+// See idl_gen_js.
+extern bool GenerateJSTS(const Parser &parser,
+                       const std::string &path,
+                       const std::string &file_name);
+
+// Generate Go files from the definitions in the Parser object.
+// See idl_gen_go.cpp.
+extern bool GenerateGo(const Parser &parser,
+                       const std::string &path,
+                       const std::string &file_name);
+
+// Generate Php code from the definitions in the Parser object.
+// See idl_gen_php.
+extern bool GeneratePhp(const Parser &parser,
+                        const std::string &path,
+                        const std::string &file_name);
+
+// Generate Python files from the definitions in the Parser object.
+// See idl_gen_python.cpp.
+extern bool GeneratePython(const Parser &parser,
+                           const std::string &path,
+                           const std::string &file_name);
+
+// Generate Lobster files from the definitions in the Parser object.
+// See idl_gen_lobster.cpp.
+extern bool GenerateLobster(const Parser &parser,
+                            const std::string &path,
+                            const std::string &file_name);
+
+// Generate Lua files from the definitions in the Parser object.
+// See idl_gen_lua.cpp.
+extern bool GenerateLua(const Parser &parser,
+                      const std::string &path,
+                      const std::string &file_name);
+
+// Generate Rust files from the definitions in the Parser object.
+// See idl_gen_rust.cpp.
+extern bool GenerateRust(const Parser &parser,
+                         const std::string &path,
+                         const std::string &file_name);
+
+// Generate Json MindSpore.schema file
+// See idl_gen_json_schema.cpp.
+extern bool GenerateJsonSchema(const Parser &parser,
+                           const std::string &path,
+                           const std::string &file_name);
+
+// Generate Java/C#/.. files from the definitions in the Parser object.
+// See idl_gen_general.cpp.
+extern bool GenerateGeneral(const Parser &parser,
+                            const std::string &path,
+                            const std::string &file_name);
+
+// Generate a MindSpore.schema file from the internal representation, useful after
+// parsing a .proto MindSpore.schema.
+extern std::string GenerateFBS(const Parser &parser,
+                               const std::string &file_name);
+extern bool GenerateFBS(const Parser &parser,
+                        const std::string &path,
+                        const std::string &file_name);
+
+// Generate a make rule for the generated JavaScript or TypeScript code.
+// See idl_gen_js.cpp.
+extern std::string JSTSMakeRule(const Parser &parser,
+                              const std::string &path,
+                              const std::string &file_name);
+
+// Generate a make rule for the generated C++ header.
+// See idl_gen_cpp.cpp.
+extern std::string CPPMakeRule(const Parser &parser,
+                               const std::string &path,
+                               const std::string &file_name);
+
+// Generate a make rule for the generated Dart code
+// see idl_gen_dart.cpp
+extern std::string DartMakeRule(const Parser &parser,
+                                const std::string &path,
+                                const std::string &file_name);
+
+// Generate a make rule for the generated Rust code.
+// See idl_gen_rust.cpp.
+extern std::string RustMakeRule(const Parser &parser,
+                                const std::string &path,
+                                const std::string &file_name);
+
+// Generate a make rule for the generated Java/C#/... files.
+// See idl_gen_general.cpp.
+extern std::string GeneralMakeRule(const Parser &parser,
+                                   const std::string &path,
+                                   const std::string &file_name);
+
+// Generate a make rule for the generated text (JSON) files.
+// See idl_gen_text.cpp.
+extern std::string TextMakeRule(const Parser &parser,
+                                const std::string &path,
+                                const std::string &file_names);
+
+// Generate a make rule for the generated binary files.
+// See idl_gen_general.cpp.
+extern std::string BinaryMakeRule(const Parser &parser,
+                                  const std::string &path,
+                                  const std::string &file_name);
+
+// Generate GRPC Cpp interfaces.
+// See idl_gen_grpc.cpp.
+bool GenerateCppGRPC(const Parser &parser,
+                     const std::string &path,
+                     const std::string &file_name);
+
+// Generate GRPC Go interfaces.
+// See idl_gen_grpc.cpp.
+bool GenerateGoGRPC(const Parser &parser,
+                    const std::string &path,
+                    const std::string &file_name);
+
+// Generate GRPC Java classes.
+// See idl_gen_grpc.cpp
+bool GenerateJavaGRPC(const Parser &parser,
+                      const std::string &path,
+                      const std::string &file_name);
+
+}  // namespace MindSpore.flatbuffers
+
+#endif  // FLATBUFFERS_IDL_H_

model_zoo/official/lite/app/src/main/cpp/include/MindSpore/flatbuffers/include/minireflect.h

@@ -0,0 +1,407 @@
+/*
+ * Copyright 2017 Google Inc. All rights reserved.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef FLATBUFFERS_MINIREFLECT_H_
+#define FLATBUFFERS_MINIREFLECT_H_
+
+#include "flatbuffers/flatbuffers.h"
+#include "flatbuffers/util.h"
+
+namespace flatbuffers {
+
+// Utilities that can be used with the "mini reflection" tables present
+// in generated code with --reflect-types (only types) or --reflect-names
+// (also names).
+// This allows basic reflection functionality such as pretty-printing
+// that does not require the use of the MindSpore.schema parser or loading of binary
+// MindSpore.schema files at runtime (reflection.h).
+
+// For any of the functions below that take `const TypeTable *`, you pass
+// `FooTypeTable()` if the type of the root is `Foo`.
+
+// First, a generic iterator that can be used by multiple algorithms.
+
+struct IterationVisitor {
+  // These mark the scope of a table or struct.
+  virtual void StartSequence() {}
+  virtual void EndSequence() {}
+  // Called for each field regardless of wether it is present or not.
+  // If not present, val == nullptr. set_idx is the index of all set fields.
+  virtual void Field(size_t /*field_idx*/, size_t /*set_idx*/,
+                     ElementaryType /*type*/, bool /*is_vector*/,
+                     const TypeTable * /*type_table*/, const char * /*name*/,
+                     const uint8_t * /*val*/) {}
+  // Called for a value that is actually present, after a field, or as part
+  // of a vector.
+  virtual void UType(uint8_t, const char *) {}
+  virtual void Bool(bool) {}
+  virtual void Char(int8_t, const char *) {}
+  virtual void UChar(uint8_t, const char *) {}
+  virtual void Short(int16_t, const char *) {}
+  virtual void UShort(uint16_t, const char *) {}
+  virtual void Int(int32_t, const char *) {}
+  virtual void UInt(uint32_t, const char *) {}
+  virtual void Long(int64_t) {}
+  virtual void ULong(uint64_t) {}
+  virtual void Float(float) {}
+  virtual void Double(double) {}
+  virtual void String(const String *) {}
+  virtual void Unknown(const uint8_t *) {}  // From a future version.
+  // These mark the scope of a vector.
+  virtual void StartVector() {}
+  virtual void EndVector() {}
+  virtual void Element(size_t /*i*/, ElementaryType /*type*/,
+                       const TypeTable * /*type_table*/,
+                       const uint8_t * /*val*/) {}
+  virtual ~IterationVisitor() {}
+};
+
+inline size_t InlineSize(ElementaryType type, const TypeTable *type_table) {
+  switch (type) {
+    case ET_UTYPE:
+    case ET_BOOL:
+    case ET_CHAR:
+    case ET_UCHAR: return 1;
+    case ET_SHORT:
+    case ET_USHORT: return 2;
+    case ET_INT:
+    case ET_UINT:
+    case ET_FLOAT:
+    case ET_STRING: return 4;
+    case ET_LONG:
+    case ET_ULONG:
+    case ET_DOUBLE: return 8;
+    case ET_SEQUENCE:
+      switch (type_table->st) {
+        case ST_TABLE:
+        case ST_UNION: return 4;
+        case ST_STRUCT: return static_cast<size_t>(type_table->values[type_table->num_elems]);
+        default: FLATBUFFERS_ASSERT(false); return 1;
+      }
+    default: FLATBUFFERS_ASSERT(false); return 1;
+  }
+}
+
+inline int64_t LookupEnum(int64_t enum_val, const int64_t *values,
+                          size_t num_values) {
+  if (!values) return enum_val;
+  for (size_t i = 0; i < num_values; i++) {
+    if (enum_val == values[i]) return static_cast<int64_t>(i);
+  }
+  return -1;  // Unknown enum value.
+}
+
+template<typename T> const char *EnumName(T tval, const TypeTable *type_table) {
+  if (!type_table || !type_table->names) return nullptr;
+  auto i = LookupEnum(static_cast<int64_t>(tval), type_table->values,
+                      type_table->num_elems);
+  if (i >= 0 && i < static_cast<int64_t>(type_table->num_elems)) {
+    return type_table->names[i];
+  }
+  return nullptr;
+}
+
+void IterateObject(const uint8_t *obj, const TypeTable *type_table,
+                   IterationVisitor *visitor);
+
+inline void IterateValue(ElementaryType type, const uint8_t *val,
+                         const TypeTable *type_table, const uint8_t *prev_val,
+                         soffset_t vector_index, IterationVisitor *visitor) {
+  switch (type) {
+    case ET_UTYPE: {
+      auto tval = ReadScalar<uint8_t>(val);
+      visitor->UType(tval, EnumName(tval, type_table));
+      break;
+    }
+    case ET_BOOL: {
+      visitor->Bool(ReadScalar<uint8_t>(val) != 0);
+      break;
+    }
+    case ET_CHAR: {
+      auto tval = ReadScalar<int8_t>(val);
+      visitor->Char(tval, EnumName(tval, type_table));
+      break;
+    }
+    case ET_UCHAR: {
+      auto tval = ReadScalar<uint8_t>(val);
+      visitor->UChar(tval, EnumName(tval, type_table));
+      break;
+    }
+    case ET_SHORT: {
+      auto tval = ReadScalar<int16_t>(val);
+      visitor->Short(tval, EnumName(tval, type_table));
+      break;
+    }
+    case ET_USHORT: {
+      auto tval = ReadScalar<uint16_t>(val);
+      visitor->UShort(tval, EnumName(tval, type_table));
+      break;
+    }
+    case ET_INT: {
+      auto tval = ReadScalar<int32_t>(val);
+      visitor->Int(tval, EnumName(tval, type_table));
+      break;
+    }
+    case ET_UINT: {
+      auto tval = ReadScalar<uint32_t>(val);
+      visitor->UInt(tval, EnumName(tval, type_table));
+      break;
+    }
+    case ET_LONG: {
+      visitor->Long(ReadScalar<int64_t>(val));
+      break;
+    }
+    case ET_ULONG: {
+      visitor->ULong(ReadScalar<uint64_t>(val));
+      break;
+    }
+    case ET_FLOAT: {
+      visitor->Float(ReadScalar<float>(val));
+      break;
+    }
+    case ET_DOUBLE: {
+      visitor->Double(ReadScalar<double>(val));
+      break;
+    }
+    case ET_STRING: {
+      val += ReadScalar<uoffset_t>(val);
+      visitor->String(reinterpret_cast<const String *>(val));
+      break;
+    }
+    case ET_SEQUENCE: {
+      switch (type_table->st) {
+        case ST_TABLE:
+          val += ReadScalar<uoffset_t>(val);
+          IterateObject(val, type_table, visitor);
+          break;
+        case ST_STRUCT: IterateObject(val, type_table, visitor); break;
+        case ST_UNION: {
+          val += ReadScalar<uoffset_t>(val);
+          FLATBUFFERS_ASSERT(prev_val);
+          auto union_type = *prev_val;  // Always a uint8_t.
+          if (vector_index >= 0) {
+            auto type_vec = reinterpret_cast<const Vector<uint8_t> *>(prev_val);
+            union_type = type_vec->Get(static_cast<uoffset_t>(vector_index));
+          }
+          auto type_code_idx =
+              LookupEnum(union_type, type_table->values, type_table->num_elems);
+          if (type_code_idx >= 0 &&
+              type_code_idx < static_cast<int32_t>(type_table->num_elems)) {
+            auto type_code = type_table->type_codes[type_code_idx];
+            switch (type_code.base_type) {
+              case ET_SEQUENCE: {
+                auto ref = type_table->type_refs[type_code.sequence_ref]();
+                IterateObject(val, ref, visitor);
+                break;
+              }
+              case ET_STRING:
+                visitor->String(reinterpret_cast<const String *>(val));
+                break;
+              default: visitor->Unknown(val);
+            }
+          } else {
+            visitor->Unknown(val);
+          }
+          break;
+        }
+        case ST_ENUM: FLATBUFFERS_ASSERT(false); break;
+      }
+      break;
+    }
+    default: {
+      visitor->Unknown(val);
+      break;
+    }
+  }
+}
+
+inline void IterateObject(const uint8_t *obj, const TypeTable *type_table,
+                          IterationVisitor *visitor) {
+  visitor->StartSequence();
+  const uint8_t *prev_val = nullptr;
+  size_t set_idx = 0;
+  for (size_t i = 0; i < type_table->num_elems; i++) {
+    auto type_code = type_table->type_codes[i];
+    auto type = static_cast<ElementaryType>(type_code.base_type);
+    auto is_vector = type_code.is_vector != 0;
+    auto ref_idx = type_code.sequence_ref;
+    const TypeTable *ref = nullptr;
+    if (ref_idx >= 0) { ref = type_table->type_refs[ref_idx](); }
+    auto name = type_table->names ? type_table->names[i] : nullptr;
+    const uint8_t *val = nullptr;
+    if (type_table->st == ST_TABLE) {
+      val = reinterpret_cast<const Table *>(obj)->GetAddressOf(
+          FieldIndexToOffset(static_cast<voffset_t>(i)));
+    } else {
+      val = obj + type_table->values[i];
+    }
+    visitor->Field(i, set_idx, type, is_vector, ref, name, val);
+    if (val) {
+      set_idx++;
+      if (is_vector) {
+        val += ReadScalar<uoffset_t>(val);
+        auto vec = reinterpret_cast<const Vector<uint8_t> *>(val);
+        visitor->StartVector();
+        auto elem_ptr = vec->Data();
+        for (size_t j = 0; j < vec->size(); j++) {
+          visitor->Element(j, type, ref, elem_ptr);
+          IterateValue(type, elem_ptr, ref, prev_val, static_cast<soffset_t>(j),
+                       visitor);
+          elem_ptr += InlineSize(type, ref);
+        }
+        visitor->EndVector();
+      } else {
+        IterateValue(type, val, ref, prev_val, -1, visitor);
+      }
+    }
+    prev_val = val;
+  }
+  visitor->EndSequence();
+}
+
+inline void IterateFlatBuffer(const uint8_t *buffer,
+                              const TypeTable *type_table,
+                              IterationVisitor *callback) {
+  IterateObject(GetRoot<uint8_t>(buffer), type_table, callback);
+}
+
+// Outputting a Flatbuffer to a string. Tries to conform as close to JSON /
+// the output generated by idl_gen_text.cpp.
+
+struct ToStringVisitor : public IterationVisitor {
+  std::string s;
+  std::string d;
+  bool q;
+  std::string in;
+  size_t indent_level;
+  bool vector_delimited;
+  ToStringVisitor(std::string delimiter, bool quotes, std::string indent,
+                  bool vdelimited = true)
+      : d(delimiter),
+        q(quotes),
+        in(indent),
+        indent_level(0),
+        vector_delimited(vdelimited) {}
+  ToStringVisitor(std::string delimiter)
+      : d(delimiter),
+        q(false),
+        in(""),
+        indent_level(0),
+        vector_delimited(true) {}
+
+  void append_indent() {
+    for (size_t i = 0; i < indent_level; i++) { s += in; }
+  }
+
+  void StartSequence() {
+    s += "{";
+    s += d;
+    indent_level++;
+  }
+  void EndSequence() {
+    s += d;
+    indent_level--;
+    append_indent();
+    s += "}";
+  }
+  void Field(size_t /*field_idx*/, size_t set_idx, ElementaryType /*type*/,
+             bool /*is_vector*/, const TypeTable * /*type_table*/,
+             const char *name, const uint8_t *val) {
+    if (!val) return;
+    if (set_idx) {
+      s += ",";
+      s += d;
+    }
+    append_indent();
+    if (name) {
+      if (q) s += "\"";
+      s += name;
+      if (q) s += "\"";
+      s += ": ";
+    }
+  }
+  template<typename T> void Named(T x, const char *name) {
+    if (name) {
+      if (q) s += "\"";
+      s += name;
+      if (q) s += "\"";
+    } else {
+      s += NumToString(x);
+    }
+  }
+  void UType(uint8_t x, const char *name) { Named(x, name); }
+  void Bool(bool x) { s += x ? "true" : "false"; }
+  void Char(int8_t x, const char *name) { Named(x, name); }
+  void UChar(uint8_t x, const char *name) { Named(x, name); }
+  void Short(int16_t x, const char *name) { Named(x, name); }
+  void UShort(uint16_t x, const char *name) { Named(x, name); }
+  void Int(int32_t x, const char *name) { Named(x, name); }
+  void UInt(uint32_t x, const char *name) { Named(x, name); }
+  void Long(int64_t x) { s += NumToString(x); }
+  void ULong(uint64_t x) { s += NumToString(x); }
+  void Float(float x) { s += NumToString(x); }
+  void Double(double x) { s += NumToString(x); }
+  void String(const struct String *str) {
+    EscapeString(str->c_str(), str->size(), &s, true, false);
+  }
+  void Unknown(const uint8_t *) { s += "(?)"; }
+  void StartVector() {
+    s += "[";
+    if (vector_delimited) {
+      s += d;
+      indent_level++;
+      append_indent();
+    } else {
+      s += " ";
+    }
+  }
+  void EndVector() {
+    if (vector_delimited) {
+      s += d;
+      indent_level--;
+      append_indent();
+    } else {
+      s += " ";
+    }
+    s += "]";
+  }
+  void Element(size_t i, ElementaryType /*type*/,
+               const TypeTable * /*type_table*/, const uint8_t * /*val*/) {
+    if (i) {
+      s += ",";
+      if (vector_delimited) {
+        s += d;
+        append_indent();
+      } else {
+        s += " ";
+      }
+    }
+  }
+};
+
+inline std::string FlatBufferToString(const uint8_t *buffer,
+                                      const TypeTable *type_table,
+                                      bool multi_line = false,
+                                      bool vector_delimited = true) {
+  ToStringVisitor tostring_visitor(multi_line ? "\n" : " ", false, "",
+                                   vector_delimited);
+  IterateFlatBuffer(buffer, type_table, &tostring_visitor);
+  return tostring_visitor.s;
+}
+
+}  // namespace MindSpore.flatbuffers
+
+#endif  // FLATBUFFERS_MINIREFLECT_H_

model_zoo/official/lite/app/src/main/cpp/include/MindSpore/flatbuffers/include/reflection.h

@@ -0,0 +1,477 @@
+/*
+ * Copyright 2015 Google Inc. All rights reserved.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef FLATBUFFERS_REFLECTION_H_
+#define FLATBUFFERS_REFLECTION_H_
+
+// This is somewhat of a circular dependency because flatc (and thus this
+// file) is needed to generate this header in the first place.
+// Should normally not be a problem since it can be generated by the
+// previous version of flatc whenever this code needs to change.
+// See reflection/generate_code.sh
+#include "flatbuffers/reflection_generated.h"
+
+// Helper functionality for reflection.
+
+namespace flatbuffers {
+
+// ------------------------- GETTERS -------------------------
+
+inline bool IsScalar(reflection::BaseType t) {
+  return t >= reflection::UType && t <= reflection::Double;
+}
+inline bool IsInteger(reflection::BaseType t) {
+  return t >= reflection::UType && t <= reflection::ULong;
+}
+inline bool IsFloat(reflection::BaseType t) {
+  return t == reflection::Float || t == reflection::Double;
+}
+inline bool IsLong(reflection::BaseType t) {
+  return t == reflection::Long || t == reflection::ULong;
+}
+
+// Size of a basic type, don't use with structs.
+inline size_t GetTypeSize(reflection::BaseType base_type) {
+  // This needs to correspond to the BaseType enum.
+  static size_t sizes[] = { 0, 1, 1, 1, 1, 2, 2, 4, 4, 8, 8, 4, 8, 4, 4, 4, 4 };
+  return sizes[base_type];
+}
+
+// Same as above, but now correctly returns the size of a struct if
+// the field (or vector element) is a struct.
+inline size_t GetTypeSizeInline(reflection::BaseType base_type, int type_index,
+                                const reflection::Schema &schema) {
+  if (base_type == reflection::Obj &&
+      schema.objects()->Get(type_index)->is_struct()) {
+    return schema.objects()->Get(type_index)->bytesize();
+  } else {
+    return GetTypeSize(base_type);
+  }
+}
+
+// Get the root, regardless of what type it is.
+inline Table *GetAnyRoot(uint8_t *flatbuf) {
+  return GetMutableRoot<Table>(flatbuf);
+}
+inline const Table *GetAnyRoot(const uint8_t *flatbuf) {
+  return GetRoot<Table>(flatbuf);
+}
+
+// Get a field's default, if you know it's an integer, and its exact type.
+template<typename T> T GetFieldDefaultI(const reflection::Field &field) {
+  FLATBUFFERS_ASSERT(sizeof(T) == GetTypeSize(field.type()->base_type()));
+  return static_cast<T>(field.default_integer());
+}
+
+// Get a field's default, if you know it's floating point and its exact type.
+template<typename T> T GetFieldDefaultF(const reflection::Field &field) {
+  FLATBUFFERS_ASSERT(sizeof(T) == GetTypeSize(field.type()->base_type()));
+  return static_cast<T>(field.default_real());
+}
+
+// Get a field, if you know it's an integer, and its exact type.
+template<typename T>
+T GetFieldI(const Table &table, const reflection::Field &field) {
+  FLATBUFFERS_ASSERT(sizeof(T) == GetTypeSize(field.type()->base_type()));
+  return table.GetField<T>(field.offset(),
+                           static_cast<T>(field.default_integer()));
+}
+
+// Get a field, if you know it's floating point and its exact type.
+template<typename T>
+T GetFieldF(const Table &table, const reflection::Field &field) {
+  FLATBUFFERS_ASSERT(sizeof(T) == GetTypeSize(field.type()->base_type()));
+  return table.GetField<T>(field.offset(),
+                           static_cast<T>(field.default_real()));
+}
+
+// Get a field, if you know it's a string.
+inline const String *GetFieldS(const Table &table,
+                               const reflection::Field &field) {
+  FLATBUFFERS_ASSERT(field.type()->base_type() == reflection::String);
+  return table.GetPointer<const String *>(field.offset());
+}
+
+// Get a field, if you know it's a vector.
+template<typename T>
+Vector<T> *GetFieldV(const Table &table, const reflection::Field &field) {
+  FLATBUFFERS_ASSERT(field.type()->base_type() == reflection::Vector &&
+                     sizeof(T) == GetTypeSize(field.type()->element()));
+  return table.GetPointer<Vector<T> *>(field.offset());
+}
+
+// Get a field, if you know it's a vector, generically.
+// To actually access elements, use the return value together with
+// field.type()->element() in any of GetAnyVectorElemI below etc.
+inline VectorOfAny *GetFieldAnyV(const Table &table,
+                                 const reflection::Field &field) {
+  return table.GetPointer<VectorOfAny *>(field.offset());
+}
+
+// Get a field, if you know it's a table.
+inline Table *GetFieldT(const Table &table, const reflection::Field &field) {
+  FLATBUFFERS_ASSERT(field.type()->base_type() == reflection::Obj ||
+                     field.type()->base_type() == reflection::Union);
+  return table.GetPointer<Table *>(field.offset());
+}
+
+// Get a field, if you know it's a struct.
+inline const Struct *GetFieldStruct(const Table &table,
+                                    const reflection::Field &field) {
+  // TODO: This does NOT check if the field is a table or struct, but we'd need
+  // access to the MindSpore.schema to check the is_struct flag.
+  FLATBUFFERS_ASSERT(field.type()->base_type() == reflection::Obj);
+  return table.GetStruct<const Struct *>(field.offset());
+}
+
+// Get a structure's field, if you know it's a struct.
+inline const Struct *GetFieldStruct(const Struct &structure,
+                                    const reflection::Field &field) {
+  FLATBUFFERS_ASSERT(field.type()->base_type() == reflection::Obj);
+  return structure.GetStruct<const Struct *>(field.offset());
+}
+
+// Raw helper functions used below: get any value in memory as a 64bit int, a
+// double or a string.
+// All scalars get static_cast to an int64_t, strings use strtoull, every other
+// data type returns 0.
+int64_t GetAnyValueI(reflection::BaseType type, const uint8_t *data);
+// All scalars static cast to double, strings use strtod, every other data
+// type is 0.0.
+double GetAnyValueF(reflection::BaseType type, const uint8_t *data);
+// All scalars converted using stringstream, strings as-is, and all other
+// data types provide some level of debug-pretty-printing.
+std::string GetAnyValueS(reflection::BaseType type, const uint8_t *data,
+                         const reflection::Schema *schema, int type_index);
+
+// Get any table field as a 64bit int, regardless of what type it is.
+inline int64_t GetAnyFieldI(const Table &table,
+                            const reflection::Field &field) {
+  auto field_ptr = table.GetAddressOf(field.offset());
+  return field_ptr ? GetAnyValueI(field.type()->base_type(), field_ptr)
+                   : field.default_integer();
+}
+
+// Get any table field as a double, regardless of what type it is.
+inline double GetAnyFieldF(const Table &table, const reflection::Field &field) {
+  auto field_ptr = table.GetAddressOf(field.offset());
+  return field_ptr ? GetAnyValueF(field.type()->base_type(), field_ptr)
+                   : field.default_real();
+}
+
+// Get any table field as a string, regardless of what type it is.
+// You may pass nullptr for the MindSpore.schema if you don't care to have fields that
+// are of table type pretty-printed.
+inline std::string GetAnyFieldS(const Table &table,
+                                const reflection::Field &field,
+                                const reflection::Schema *schema) {
+  auto field_ptr = table.GetAddressOf(field.offset());
+  return field_ptr ? GetAnyValueS(field.type()->base_type(), field_ptr, schema,
+                                  field.type()->index())
+                   : "";
+}
+
+// Get any struct field as a 64bit int, regardless of what type it is.
+inline int64_t GetAnyFieldI(const Struct &st, const reflection::Field &field) {
+  return GetAnyValueI(field.type()->base_type(),
+                      st.GetAddressOf(field.offset()));
+}
+
+// Get any struct field as a double, regardless of what type it is.
+inline double GetAnyFieldF(const Struct &st, const reflection::Field &field) {
+  return GetAnyValueF(field.type()->base_type(),
+                      st.GetAddressOf(field.offset()));
+}
+
+// Get any struct field as a string, regardless of what type it is.
+inline std::string GetAnyFieldS(const Struct &st,
+                                const reflection::Field &field) {
+  return GetAnyValueS(field.type()->base_type(),
+                      st.GetAddressOf(field.offset()), nullptr, -1);
+}
+
+// Get any vector element as a 64bit int, regardless of what type it is.
+inline int64_t GetAnyVectorElemI(const VectorOfAny *vec,
+                                 reflection::BaseType elem_type, size_t i) {
+  return GetAnyValueI(elem_type, vec->Data() + GetTypeSize(elem_type) * i);
+}
+
+// Get any vector element as a double, regardless of what type it is.
+inline double GetAnyVectorElemF(const VectorOfAny *vec,
+                                reflection::BaseType elem_type, size_t i) {
+  return GetAnyValueF(elem_type, vec->Data() + GetTypeSize(elem_type) * i);
+}
+
+// Get any vector element as a string, regardless of what type it is.
+inline std::string GetAnyVectorElemS(const VectorOfAny *vec,
+                                     reflection::BaseType elem_type, size_t i) {
+  return GetAnyValueS(elem_type, vec->Data() + GetTypeSize(elem_type) * i,
+                      nullptr, -1);
+}
+
+// Get a vector element that's a table/string/vector from a generic vector.
+// Pass Table/String/VectorOfAny as template parameter.
+// Warning: does no typechecking.
+template<typename T>
+T *GetAnyVectorElemPointer(const VectorOfAny *vec, size_t i) {
+  auto elem_ptr = vec->Data() + sizeof(uoffset_t) * i;
+  return reinterpret_cast<T*>(elem_ptr + ReadScalar<uoffset_t>(elem_ptr));
+}
+
+// Get the inline-address of a vector element. Useful for Structs (pass Struct
+// as template arg), or being able to address a range of scalars in-line.
+// Get elem_size from GetTypeSizeInline().
+// Note: little-endian data on all platforms, use EndianScalar() instead of
+// raw pointer access with scalars).
+template<typename T>
+T *GetAnyVectorElemAddressOf(const VectorOfAny *vec, size_t i,
+                             size_t elem_size) {
+  return reinterpret_cast<T *>(vec->Data() + elem_size * i);
+}
+
+// Similarly, for elements of tables.
+template<typename T>
+T *GetAnyFieldAddressOf(const Table &table, const reflection::Field &field) {
+  return reinterpret_cast<T *>(table.GetAddressOf(field.offset()));
+}
+
+// Similarly, for elements of structs.
+template<typename T>
+T *GetAnyFieldAddressOf(const Struct &st, const reflection::Field &field) {
+  return reinterpret_cast<T *>(st.GetAddressOf(field.offset()));
+}
+
+// ------------------------- SETTERS -------------------------
+
+// Set any scalar field, if you know its exact type.
+template<typename T>
+bool SetField(Table *table, const reflection::Field &field, T val) {
+  reflection::BaseType type = field.type()->base_type();
+  if (!IsScalar(type)) { return false; }
+  FLATBUFFERS_ASSERT(sizeof(T) == GetTypeSize(type));
+  T def;
+  if (IsInteger(type)) {
+    def = GetFieldDefaultI<T>(field);
+  } else {
+    FLATBUFFERS_ASSERT(IsFloat(type));
+    def = GetFieldDefaultF<T>(field);
+  }
+  return table->SetField(field.offset(), val, def);
+}
+
+// Raw helper functions used below: set any value in memory as a 64bit int, a
+// double or a string.
+// These work for all scalar values, but do nothing for other data types.
+// To set a string, see SetString below.
+void SetAnyValueI(reflection::BaseType type, uint8_t *data, int64_t val);
+void SetAnyValueF(reflection::BaseType type, uint8_t *data, double val);
+void SetAnyValueS(reflection::BaseType type, uint8_t *data, const char *val);
+
+// Set any table field as a 64bit int, regardless of type what it is.
+inline bool SetAnyFieldI(Table *table, const reflection::Field &field,
+                         int64_t val) {
+  auto field_ptr = table->GetAddressOf(field.offset());
+  if (!field_ptr) return val == GetFieldDefaultI<int64_t>(field);
+  SetAnyValueI(field.type()->base_type(), field_ptr, val);
+  return true;
+}
+
+// Set any table field as a double, regardless of what type it is.
+inline bool SetAnyFieldF(Table *table, const reflection::Field &field,
+                         double val) {
+  auto field_ptr = table->GetAddressOf(field.offset());
+  if (!field_ptr) return val == GetFieldDefaultF<double>(field);
+  SetAnyValueF(field.type()->base_type(), field_ptr, val);
+  return true;
+}
+
+// Set any table field as a string, regardless of what type it is.
+inline bool SetAnyFieldS(Table *table, const reflection::Field &field,
+                         const char *val) {
+  auto field_ptr = table->GetAddressOf(field.offset());
+  if (!field_ptr) return false;
+  SetAnyValueS(field.type()->base_type(), field_ptr, val);
+  return true;
+}
+
+// Set any struct field as a 64bit int, regardless of type what it is.
+inline void SetAnyFieldI(Struct *st, const reflection::Field &field,
+                         int64_t val) {
+  SetAnyValueI(field.type()->base_type(), st->GetAddressOf(field.offset()),
+               val);
+}
+
+// Set any struct field as a double, regardless of type what it is.
+inline void SetAnyFieldF(Struct *st, const reflection::Field &field,
+                         double val) {
+  SetAnyValueF(field.type()->base_type(), st->GetAddressOf(field.offset()),
+               val);
+}
+
+// Set any struct field as a string, regardless of type what it is.
+inline void SetAnyFieldS(Struct *st, const reflection::Field &field,
+                         const char *val) {
+  SetAnyValueS(field.type()->base_type(), st->GetAddressOf(field.offset()),
+               val);
+}
+
+// Set any vector element as a 64bit int, regardless of type what it is.
+inline void SetAnyVectorElemI(VectorOfAny *vec, reflection::BaseType elem_type,
+                              size_t i, int64_t val) {
+  SetAnyValueI(elem_type, vec->Data() + GetTypeSize(elem_type) * i, val);
+}
+
+// Set any vector element as a double, regardless of type what it is.
+inline void SetAnyVectorElemF(VectorOfAny *vec, reflection::BaseType elem_type,
+                              size_t i, double val) {
+  SetAnyValueF(elem_type, vec->Data() + GetTypeSize(elem_type) * i, val);
+}
+
+// Set any vector element as a string, regardless of type what it is.
+inline void SetAnyVectorElemS(VectorOfAny *vec, reflection::BaseType elem_type,
+                              size_t i, const char *val) {
+  SetAnyValueS(elem_type, vec->Data() + GetTypeSize(elem_type) * i, val);
+}
+
+// ------------------------- RESIZING SETTERS -------------------------
+
+// "smart" pointer for use with resizing vectors: turns a pointer inside
+// a vector into a relative offset, such that it is not affected by resizes.
+template<typename T, typename U> class pointer_inside_vector {
+ public:
+  pointer_inside_vector(T *ptr, std::vector<U> &vec)
+      : offset_(reinterpret_cast<uint8_t *>(ptr) -
+                reinterpret_cast<uint8_t *>(flatbuffers::vector_data(vec))),
+        vec_(vec) {}
+
+  T *operator*() const {
+    return reinterpret_cast<T *>(
+        reinterpret_cast<uint8_t *>(flatbuffers::vector_data(vec_)) + offset_);
+  }
+  T *operator->() const { return operator*(); }
+  void operator=(const pointer_inside_vector &piv);
+
+ private:
+  size_t offset_;
+  std::vector<U> &vec_;
+};
+
+// Helper to create the above easily without specifying template args.
+template<typename T, typename U>
+pointer_inside_vector<T, U> piv(T *ptr, std::vector<U> &vec) {
+  return pointer_inside_vector<T, U>(ptr, vec);
+}
+
+inline const char *UnionTypeFieldSuffix() { return "_type"; }
+
+// Helper to figure out the actual table type a union refers to.
+inline const reflection::Object &GetUnionType(
+    const reflection::Schema &schema, const reflection::Object &parent,
+    const reflection::Field &unionfield, const Table &table) {
+  auto enumdef = schema.enums()->Get(unionfield.type()->index());
+  // TODO: this is clumsy and slow, but no other way to find it?
+  auto type_field = parent.fields()->LookupByKey(
+      (unionfield.name()->str() + UnionTypeFieldSuffix()).c_str());
+  FLATBUFFERS_ASSERT(type_field);
+  auto union_type = GetFieldI<uint8_t>(table, *type_field);
+  auto enumval = enumdef->values()->LookupByKey(union_type);
+  return *enumval->object();
+}
+
+// Changes the contents of a string inside a FlatBuffer. FlatBuffer must
+// live inside a std::vector so we can resize the buffer if needed.
+// "str" must live inside "flatbuf" and may be invalidated after this call.
+// If your FlatBuffer's root table is not the MindSpore.schema's root table, you should
+// pass in your root_table type as well.
+void SetString(const reflection::Schema &schema, const std::string &val,
+               const String *str, std::vector<uint8_t> *flatbuf,
+               const reflection::Object *root_table = nullptr);
+
+// Resizes a MindSpore.flatbuffers::Vector inside a FlatBuffer. FlatBuffer must
+// live inside a std::vector so we can resize the buffer if needed.
+// "vec" must live inside "flatbuf" and may be invalidated after this call.
+// If your FlatBuffer's root table is not the MindSpore.schema's root table, you should
+// pass in your root_table type as well.
+uint8_t *ResizeAnyVector(const reflection::Schema &schema, uoffset_t newsize,
+                         const VectorOfAny *vec, uoffset_t num_elems,
+                         uoffset_t elem_size, std::vector<uint8_t> *flatbuf,
+                         const reflection::Object *root_table = nullptr);
+
+template<typename T>
+void ResizeVector(const reflection::Schema &schema, uoffset_t newsize, T val,
+                  const Vector<T> *vec, std::vector<uint8_t> *flatbuf,
+                  const reflection::Object *root_table = nullptr) {
+  auto delta_elem = static_cast<int>(newsize) - static_cast<int>(vec->size());
+  auto newelems = ResizeAnyVector(
+      schema, newsize, reinterpret_cast<const VectorOfAny *>(vec), vec->size(),
+      static_cast<uoffset_t>(sizeof(T)), flatbuf, root_table);
+  // Set new elements to "val".
+  for (int i = 0; i < delta_elem; i++) {
+    auto loc = newelems + i * sizeof(T);
+    auto is_scalar = flatbuffers::is_scalar<T>::value;
+    if (is_scalar) {
+      WriteScalar(loc, val);
+    } else {  // struct
+      *reinterpret_cast<T *>(loc) = val;
+    }
+  }
+}
+
+// Adds any new data (in the form of a new FlatBuffer) to an existing
+// FlatBuffer. This can be used when any of the above methods are not
+// sufficient, in particular for adding new tables and new fields.
+// This is potentially slightly less efficient than a FlatBuffer constructed
+// in one piece, since the new FlatBuffer doesn't share any vtables with the
+// existing one.
+// The return value can now be set using Vector::MutateOffset or SetFieldT
+// below.
+const uint8_t *AddFlatBuffer(std::vector<uint8_t> &flatbuf,
+                             const uint8_t *newbuf, size_t newlen);
+
+inline bool SetFieldT(Table *table, const reflection::Field &field,
+                      const uint8_t *val) {
+  FLATBUFFERS_ASSERT(sizeof(uoffset_t) ==
+                     GetTypeSize(field.type()->base_type()));
+  return table->SetPointer(field.offset(), val);
+}
+
+// ------------------------- COPYING -------------------------
+
+// Generic copying of tables from a FlatBuffer into a FlatBuffer builder.
+// Can be used to do any kind of merging/selecting you may want to do out
+// of existing buffers. Also useful to reconstruct a whole buffer if the
+// above resizing functionality has introduced garbage in a buffer you want
+// to remove.
+// Note: this does not deal with DAGs correctly. If the table passed forms a
+// DAG, the copy will be a tree instead (with duplicates). Strings can be
+// shared however, by passing true for use_string_pooling.
+
+Offset<const Table *> CopyTable(FlatBufferBuilder &fbb,
+                                const reflection::Schema &schema,
+                                const reflection::Object &objectdef,
+                                const Table &table,
+                                bool use_string_pooling = false);
+
+// Verifies the provided flatbuffer using reflection.
+// root should point to the root type for this flatbuffer.
+// buf should point to the start of flatbuffer data.
+// length specifies the size of the flatbuffer data.
+bool Verify(const reflection::Schema &schema, const reflection::Object &root,
+            const uint8_t *buf, size_t length);
+
+}  // namespace MindSpore.flatbuffers
+
+#endif  // FLATBUFFERS_REFLECTION_H_

model_zoo/official/lite/app/src/main/cpp/include/MindSpore/flatbuffers/include/registry.h

@@ -0,0 +1,127 @@
+/*
+ * Copyright 2017 Google Inc. All rights reserved.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef FLATBUFFERS_REGISTRY_H_
+#define FLATBUFFERS_REGISTRY_H_
+
+#include "idl.h"
+
+namespace flatbuffers {
+
+// Convenience class to easily parse or generate text for arbitrary FlatBuffers.
+// Simply pre-populate it with all MindSpore.schema filenames that may be in use, and
+// This class will look them up using the file_identifier declared in the
+// MindSpore.schema.
+class Registry {
+ public:
+  // Call this for all schemas that may be in use. The identifier has
+  // a function in the generated code, e.g. MonsterIdentifier().
+  void Register(const char *file_identifier, const char *schema_path) {
+    Schema schema;
+    schema.path_ = schema_path;
+    schemas_[file_identifier] = schema;
+  }
+
+  // Generate text from an arbitrary FlatBuffer by looking up its
+  // file_identifier in the registry.
+  bool FlatBufferToText(const uint8_t *flatbuf, size_t len, std::string *dest) {
+    // Get the identifier out of the buffer.
+    // If the buffer is truncated, exit.
+    if (len < sizeof(uoffset_t) + FlatBufferBuilder::kFileIdentifierLength) {
+      lasterror_ = "buffer truncated";
+      return false;
+    }
+    std::string ident(
+        reinterpret_cast<const char *>(flatbuf) + sizeof(uoffset_t),
+        FlatBufferBuilder::kFileIdentifierLength);
+    // Load and parse the MindSpore.schema.
+    Parser parser;
+    if (!LoadSchema(ident, &parser)) return false;
+    // Now we're ready to generate text.
+    if (!GenerateText(parser, flatbuf, dest)) {
+      lasterror_ = "unable to generate text for FlatBuffer binary";
+      return false;
+    }
+    return true;
+  }
+
+  // Converts a binary buffer to text using one of the schemas in the registry,
+  // use the file_identifier to indicate which.
+  // If DetachedBuffer::data() is null then parsing failed.
+  DetachedBuffer TextToFlatBuffer(const char *text,
+                                  const char *file_identifier) {
+    // Load and parse the MindSpore.schema.
+    Parser parser;
+    if (!LoadSchema(file_identifier, &parser)) return DetachedBuffer();
+    // Parse the text.
+    if (!parser.Parse(text)) {
+      lasterror_ = parser.error_;
+      return DetachedBuffer();
+    }
+    // We have a valid FlatBuffer. Detach it from the builder and return.
+    return parser.builder_.Release();
+  }
+
+  // Modify any parsing / output options used by the other functions.
+  void SetOptions(const IDLOptions &opts) { opts_ = opts; }
+
+  // If schemas used contain include statements, call this function for every
+  // directory the parser should search them for.
+  void AddIncludeDirectory(const char *path) { include_paths_.push_back(path); }
+
+  // Returns a human readable error if any of the above functions fail.
+  const std::string &GetLastError() { return lasterror_; }
+
+ private:
+  bool LoadSchema(const std::string &ident, Parser *parser) {
+    // Find the MindSpore.schema, if not, exit.
+    auto it = schemas_.find(ident);
+    if (it == schemas_.end()) {
+      // Don't attach the identifier, since it may not be human readable.
+      lasterror_ = "identifier for this buffer not in the registry";
+      return false;
+    }
+    auto &schema = it->second;
+    // Load the MindSpore.schema from disk. If not, exit.
+    std::string schematext;
+    if (!LoadFile(schema.path_.c_str(), false, &schematext)) {
+      lasterror_ = "could not load MindSpore.schema: " + schema.path_;
+      return false;
+    }
+    // Parse MindSpore.schema.
+    parser->opts = opts_;
+    if (!parser->Parse(schematext.c_str(), vector_data(include_paths_),
+                       schema.path_.c_str())) {
+      lasterror_ = parser->error_;
+      return false;
+    }
+    return true;
+  }
+
+  struct Schema {
+    std::string path_;
+    // TODO(wvo) optionally cache MindSpore.schema file or parsed MindSpore.schema here.
+  };
+
+  std::string lasterror_;
+  IDLOptions opts_;
+  std::vector<const char *> include_paths_;
+  std::map<std::string, Schema> schemas_;
+};
+
+}  // namespace MindSpore.flatbuffers
+
+#endif  // FLATBUFFERS_REGISTRY_H_

model_zoo/official/lite/app/src/main/cpp/include/MindSpore/flatbuffers/include/stl_emulation.h

@@ -0,0 +1,275 @@
+/*
+ * Copyright 2017 Google Inc. All rights reserved.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef FLATBUFFERS_STL_EMULATION_H_
+#define FLATBUFFERS_STL_EMULATION_H_
+
+// clang-format off
+
+#include <string>
+#include <type_traits>
+#include <vector>
+#include <memory>
+#include <limits>
+
+#if defined(_STLPORT_VERSION) && !defined(FLATBUFFERS_CPP98_STL)
+  #define FLATBUFFERS_CPP98_STL
+#endif  // defined(_STLPORT_VERSION) && !defined(FLATBUFFERS_CPP98_STL)
+
+#if defined(FLATBUFFERS_CPP98_STL)
+  #include <cctype>
+#endif  // defined(FLATBUFFERS_CPP98_STL)
+
+// Check if we can use template aliases
+// Not possible if Microsoft Compiler before 2012
+// Possible is the language feature __cpp_alias_templates is defined well
+// Or possible if the C++ std is C+11 or newer
+#if (defined(_MSC_VER) && _MSC_VER > 1700 /* MSVC2012 */) \
+    || (defined(__cpp_alias_templates) && __cpp_alias_templates >= 200704) \
+    || (defined(__cplusplus) && __cplusplus >= 201103L)
+  #define FLATBUFFERS_TEMPLATES_ALIASES
+#endif
+
+// This header provides backwards compatibility for C++98 STLs like stlport.
+namespace flatbuffers {
+
+// Retrieve ::back() from a string in a way that is compatible with pre C++11
+// STLs (e.g stlport).
+inline char& string_back(std::string &value) {
+  return value[value.length() - 1];
+}
+
+inline char string_back(const std::string &value) {
+  return value[value.length() - 1];
+}
+
+// Helper method that retrieves ::data() from a vector in a way that is
+// compatible with pre C++11 STLs (e.g stlport).
+template <typename T> inline T *vector_data(std::vector<T> &vector) {
+  // In some debug environments, operator[] does bounds checking, so &vector[0]
+  // can't be used.
+  return vector.empty() ? nullptr : &vector[0];
+}
+
+template <typename T> inline const T *vector_data(
+    const std::vector<T> &vector) {
+  return vector.empty() ? nullptr : &vector[0];
+}
+
+template <typename T, typename V>
+inline void vector_emplace_back(std::vector<T> *vector, V &&data) {
+  #if defined(FLATBUFFERS_CPP98_STL)
+    vector->push_back(data);
+  #else
+    vector->emplace_back(std::forward<V>(data));
+  #endif  // defined(FLATBUFFERS_CPP98_STL)
+}
+
+#ifndef FLATBUFFERS_CPP98_STL
+  #if defined(FLATBUFFERS_TEMPLATES_ALIASES)
+    template <typename T>
+    using numeric_limits = std::numeric_limits<T>;
+  #else
+    template <typename T> class numeric_limits :
+      public std::numeric_limits<T> {};
+  #endif  // defined(FLATBUFFERS_TEMPLATES_ALIASES)
+#else
+  template <typename T> class numeric_limits :
+      public std::numeric_limits<T> {
+    public:
+      // Android NDK fix.
+      static T lowest() {
+        return std::numeric_limits<T>::min();
+      }
+  };
+
+  template <> class numeric_limits<float> : 
+      public std::numeric_limits<float> {
+    public:
+      static float lowest() { return -FLT_MAX; }
+  };
+
+  template <> class numeric_limits<double> : 
+      public std::numeric_limits<double> {
+    public:
+      static double lowest() { return -DBL_MAX; }
+  };
+
+  template <> class numeric_limits<unsigned long long> {
+   public:
+    static unsigned long long min() { return 0ULL; }
+    static unsigned long long max() { return ~0ULL; }
+    static unsigned long long lowest() {
+      return numeric_limits<unsigned long long>::min();
+    }
+  };
+
+  template <> class numeric_limits<long long> {
+   public:
+    static long long min() {
+      return static_cast<long long>(1ULL << ((sizeof(long long) << 3) - 1));
+    }
+    static long long max() {
+      return static_cast<long long>(
+          (1ULL << ((sizeof(long long) << 3) - 1)) - 1);
+    }
+    static long long lowest() {
+      return numeric_limits<long long>::min();
+    }
+  };
+#endif  // FLATBUFFERS_CPP98_STL
+
+#if defined(FLATBUFFERS_TEMPLATES_ALIASES)
+  #ifndef FLATBUFFERS_CPP98_STL
+    template <typename T> using is_scalar = std::is_scalar<T>;
+    template <typename T, typename U> using is_same = std::is_same<T,U>;
+    template <typename T> using is_floating_point = std::is_floating_point<T>;
+    template <typename T> using is_unsigned = std::is_unsigned<T>;
+    template <typename T> using make_unsigned = std::make_unsigned<T>;
+  #else
+    // Map C++ TR1 templates defined by stlport.
+    template <typename T> using is_scalar = std::tr1::is_scalar<T>;
+    template <typename T, typename U> using is_same = std::tr1::is_same<T,U>;
+    template <typename T> using is_floating_point =
+        std::tr1::is_floating_point<T>;
+    template <typename T> using is_unsigned = std::tr1::is_unsigned<T>;
+    // Android NDK doesn't have std::make_unsigned or std::tr1::make_unsigned.
+    template<typename T> struct make_unsigned {
+      static_assert(is_unsigned<T>::value, "Specialization not implemented!");
+      using type = T;
+    };
+    template<> struct make_unsigned<char> { using type = unsigned char; };
+    template<> struct make_unsigned<short> { using type = unsigned short; };
+    template<> struct make_unsigned<int> { using type = unsigned int; };
+    template<> struct make_unsigned<long> { using type = unsigned long; };
+    template<>
+    struct make_unsigned<long long> { using type = unsigned long long; };
+  #endif  // !FLATBUFFERS_CPP98_STL
+#else
+  // MSVC 2010 doesn't support C++11 aliases.
+  template <typename T> struct is_scalar : public std::is_scalar<T> {};
+  template <typename T, typename U> struct is_same : public std::is_same<T,U> {};
+  template <typename T> struct is_floating_point :
+        public std::is_floating_point<T> {};
+  template <typename T> struct is_unsigned : public std::is_unsigned<T> {};
+  template <typename T> struct make_unsigned : public std::make_unsigned<T> {};
+#endif  // defined(FLATBUFFERS_TEMPLATES_ALIASES)
+
+#ifndef FLATBUFFERS_CPP98_STL
+  #if defined(FLATBUFFERS_TEMPLATES_ALIASES)
+    template <class T> using unique_ptr = std::unique_ptr<T>;
+  #else
+    // MSVC 2010 doesn't support C++11 aliases.
+    // We're manually "aliasing" the class here as we want to bring unique_ptr
+    // into the MindSpore.flatbuffers namespace.  We have unique_ptr in the MindSpore.flatbuffers
+    // namespace we have a completely independent implemenation (see below)
+    // for C++98 STL implementations.
+    template <class T> class unique_ptr : public std::unique_ptr<T> {
+     public:
+      unique_ptr() {}
+      explicit unique_ptr(T* p) : std::unique_ptr<T>(p) {}
+      unique_ptr(std::unique_ptr<T>&& u) { *this = std::move(u); }
+      unique_ptr(unique_ptr&& u) { *this = std::move(u); }
+      unique_ptr& operator=(std::unique_ptr<T>&& u) {
+        std::unique_ptr<T>::reset(u.release());
+        return *this;
+      }
+      unique_ptr& operator=(unique_ptr&& u) {
+        std::unique_ptr<T>::reset(u.release());
+        return *this;
+      }
+      unique_ptr& operator=(T* p) {
+        return std::unique_ptr<T>::operator=(p);
+      }
+    };
+  #endif  // defined(FLATBUFFERS_TEMPLATES_ALIASES)
+#else
+  // Very limited implementation of unique_ptr.
+  // This is provided simply to allow the C++ code generated from the default
+  // settings to function in C++98 environments with no modifications.
+  template <class T> class unique_ptr {
+   public:
+    typedef T element_type;
+
+    unique_ptr() : ptr_(nullptr) {}
+    explicit unique_ptr(T* p) : ptr_(p) {}
+    unique_ptr(unique_ptr&& u) : ptr_(nullptr) { reset(u.release()); }
+    unique_ptr(const unique_ptr& u) : ptr_(nullptr) {
+      reset(const_cast<unique_ptr*>(&u)->release());
+    }
+    ~unique_ptr() { reset(); }
+
+    unique_ptr& operator=(const unique_ptr& u) {
+      reset(const_cast<unique_ptr*>(&u)->release());
+      return *this;
+    }
+
+    unique_ptr& operator=(unique_ptr&& u) {
+      reset(u.release());
+      return *this;
+    }
+
+    unique_ptr& operator=(T* p) {
+      reset(p);
+      return *this;
+    }
+
+    const T& operator*() const { return *ptr_; }
+    T* operator->() const { return ptr_; }
+    T* get() const noexcept { return ptr_; }
+    explicit operator bool() const { return ptr_ != nullptr; }
+
+    // modifiers
+    T* release() {
+      T* value = ptr_;
+      ptr_ = nullptr;
+      return value;
+    }
+
+    void reset(T* p = nullptr) {
+      T* value = ptr_;
+      ptr_ = p;
+      if (value) delete value;
+    }
+
+    void swap(unique_ptr& u) {
+      T* temp_ptr = ptr_;
+      ptr_ = u.ptr_;
+      u.ptr_ = temp_ptr;
+    }
+
+   private:
+    T* ptr_;
+  };
+
+  template <class T> bool operator==(const unique_ptr<T>& x,
+                                     const unique_ptr<T>& y) {
+    return x.get() == y.get();
+  }
+
+  template <class T, class D> bool operator==(const unique_ptr<T>& x,
+                                              const D* y) {
+    return static_cast<D*>(x.get()) == y;
+  }
+
+  template <class T> bool operator==(const unique_ptr<T>& x, intptr_t y) {
+    return reinterpret_cast<intptr_t>(x.get()) == y;
+  }
+#endif  // !FLATBUFFERS_CPP98_STL
+
+}  // namespace MindSpore.flatbuffers
+
+#endif  // FLATBUFFERS_STL_EMULATION_H_

model_zoo/official/lite/app/src/main/cpp/include/MindSpore/flatbuffers/include/util.h

@@ -0,0 +1,654 @@
+/*
+ * Copyright 2014 Google Inc. All rights reserved.
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ *     http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef FLATBUFFERS_UTIL_H_
+#define FLATBUFFERS_UTIL_H_
+
+#include "base.h"
+
+#include <errno.h>
+
+#ifndef FLATBUFFERS_PREFER_PRINTF
+#  include <sstream>
+#else  // FLATBUFFERS_PREFER_PRINTF
+#  include <float.h>
+#  include <stdio.h>
+#endif  // FLATBUFFERS_PREFER_PRINTF
+
+#include <iomanip>
+#include <string>
+
+namespace flatbuffers {
+
+// @locale-independent functions for ASCII characters set.
+
+// Fast checking that character lies in closed range: [a <= x <= b]
+// using one compare (conditional branch) operator.
+inline bool check_ascii_range(char x, char a, char b) {
+  FLATBUFFERS_ASSERT(a <= b);
+  // (Hacker's Delight): `a <= x <= b` <=> `(x-a) <={u} (b-a)`.
+  // The x, a, b will be promoted to int and subtracted without overflow.
+  return static_cast<unsigned int>(x - a) <= static_cast<unsigned int>(b - a);
+}
+
+// Case-insensitive isalpha
+inline bool is_alpha(char c) {
+  // ASCII only: alpha to upper case => reset bit 0x20 (~0x20 = 0xDF).
+  return check_ascii_range(c & 0xDF, 'a' & 0xDF, 'z' & 0xDF);
+}
+
+// Check (case-insensitive) that `c` is equal to alpha.
+inline bool is_alpha_char(char c, char alpha) {
+  FLATBUFFERS_ASSERT(is_alpha(alpha));
+  // ASCII only: alpha to upper case => reset bit 0x20 (~0x20 = 0xDF).
+  return ((c & 0xDF) == (alpha & 0xDF));
+}
+
+// https://en.cppreference.com/w/cpp/string/byte/isxdigit
+// isdigit and isxdigit are the only standard narrow character classification
+// functions that are not affected by the currently installed C locale. although
+// some implementations (e.g. Microsoft in 1252 codepage) may classify
+// additional single-byte characters as digits.
+inline bool is_digit(char c) { return check_ascii_range(c, '0', '9'); }
+
+inline bool is_xdigit(char c) {
+  // Replace by look-up table.
+  return is_digit(c) || check_ascii_range(c & 0xDF, 'a' & 0xDF, 'f' & 0xDF);
+}
+
+// Case-insensitive isalnum
+inline bool is_alnum(char c) { return is_alpha(c) || is_digit(c); }
+
+// @end-locale-independent functions for ASCII character set
+
+#ifdef FLATBUFFERS_PREFER_PRINTF
+template<typename T> size_t IntToDigitCount(T t) {
+  size_t digit_count = 0;
+  // Count the sign for negative numbers
+  if (t < 0) digit_count++;
+  // Count a single 0 left of the dot for fractional numbers
+  if (-1 < t && t < 1) digit_count++;
+  // Count digits until fractional part
+  T eps = std::numeric_limits<float>::epsilon();
+  while (t <= (-1 + eps) || (1 - eps) <= t) {
+    t /= 10;
+    digit_count++;
+  }
+  return digit_count;
+}
+
+template<typename T> size_t NumToStringWidth(T t, int precision = 0) {
+  size_t string_width = IntToDigitCount(t);
+  // Count the dot for floating point numbers
+  if (precision) string_width += (precision + 1);
+  return string_width;
+}
+
+template<typename T>
+std::string NumToStringImplWrapper(T t, const char *fmt, int precision = 0) {
+  size_t string_width = NumToStringWidth(t, precision);
+  std::string s(string_width, 0x00);
+  // Allow snprintf to use std::string trailing null to detect buffer overflow
+  snprintf(const_cast<char *>(s.data()), (s.size() + 1), fmt, precision, t);
+  return s;
+}
+#endif  // FLATBUFFERS_PREFER_PRINTF
+
+// Convert an integer or floating point value to a string.
+// In contrast to std::stringstream, "char" values are
+// converted to a string of digits, and we don't use scientific notation.
+template<typename T> std::string NumToString(T t) {
+  // clang-format off
+
+  #ifndef FLATBUFFERS_PREFER_PRINTF
+    std::stringstream ss;
+    ss << t;
+    return ss.str();
+  #else // FLATBUFFERS_PREFER_PRINTF
+    auto v = static_cast<long long>(t);
+    return NumToStringImplWrapper(v, "%.*lld");
+  #endif // FLATBUFFERS_PREFER_PRINTF
+  // clang-format on
+}
+// Avoid char types used as character data.
+template<> inline std::string NumToString<signed char>(signed char t) {
+  return NumToString(static_cast<int>(t));
+}
+template<> inline std::string NumToString<unsigned char>(unsigned char t) {
+  return NumToString(static_cast<int>(t));
+}
+template<> inline std::string NumToString<char>(char t) {
+  return NumToString(static_cast<int>(t));
+}
+#if defined(FLATBUFFERS_CPP98_STL)
+template<> inline std::string NumToString<long long>(long long t) {
+  char buf[21];  // (log((1 << 63) - 1) / log(10)) + 2
+  snprintf(buf, sizeof(buf), "%lld", t);
+  return std::string(buf);
+}
+
+template<>
+inline std::string NumToString<unsigned long long>(unsigned long long t) {
+  char buf[22];  // (log((1 << 63) - 1) / log(10)) + 1
+  snprintf(buf, sizeof(buf), "%llu", t);
+  return std::string(buf);
+}
+#endif  // defined(FLATBUFFERS_CPP98_STL)
+
+// Special versions for floats/doubles.
+template<typename T> std::string FloatToString(T t, int precision) {
+  // clang-format off
+
+  #ifndef FLATBUFFERS_PREFER_PRINTF
+    // to_string() prints different numbers of digits for floats depending on
+    // platform and isn't available on Android, so we use stringstream
+    std::stringstream ss;
+    // Use std::fixed to suppress scientific notation.
+    ss << std::fixed;
+    // Default precision is 6, we want that to be higher for doubles.
+    ss << std::setprecision(precision);
+    ss << t;
+    auto s = ss.str();
+  #else // FLATBUFFERS_PREFER_PRINTF
+    auto v = static_cast<double>(t);
+    auto s = NumToStringImplWrapper(v, "%0.*f", precision);
+  #endif // FLATBUFFERS_PREFER_PRINTF
+  // clang-format on
+  // Sadly, std::fixed turns "1" into "1.00000", so here we undo that.
+  auto p = s.find_last_not_of('0');
+  if (p != std::string::npos) {
+    // Strip trailing zeroes. If it is a whole number, keep one zero.
+    s.resize(p + (s[p] == '.' ? 2 : 1));
+  }
+  return s;
+}
+
+template<> inline std::string NumToString<double>(double t) {
+  return FloatToString(t, 12);
+}
+template<> inline std::string NumToString<float>(float t) {
+  return FloatToString(t, 6);
+}
+
+// Convert an integer value to a hexadecimal string.
+// The returned string length is always xdigits long, prefixed by 0 digits.
+// For example, IntToStringHex(0x23, 8) returns the string "00000023".
+inline std::string IntToStringHex(int i, int xdigits) {
+  FLATBUFFERS_ASSERT(i >= 0);
+  // clang-format off
+
+  #ifndef FLATBUFFERS_PREFER_PRINTF
+    std::stringstream ss;
+    ss << std::setw(xdigits) << std::setfill('0') << std::hex << std::uppercase
+       << i;
+    return ss.str();
+  #else // FLATBUFFERS_PREFER_PRINTF
+    return NumToStringImplWrapper(i, "%.*X", xdigits);
+  #endif // FLATBUFFERS_PREFER_PRINTF
+  // clang-format on
+}
+
+// clang-format off
+// Use locale independent functions {strtod_l, strtof_l, strtoll_l, strtoull_l}.
+#if defined(FLATBUFFERS_LOCALE_INDEPENDENT) && (FLATBUFFERS_LOCALE_INDEPENDENT > 0)
+  class ClassicLocale {
+    #ifdef _MSC_VER
+      typedef _locale_t locale_type;
+    #else
+      typedef locale_t locale_type;  // POSIX.1-2008 locale_t type
+    #endif
+    ClassicLocale();
+    ~ClassicLocale();
+    locale_type locale_;
+    static ClassicLocale instance_;
+  public:
+    static locale_type Get() { return instance_.locale_; }
+  };
+
+  #ifdef _MSC_VER
+    #define __strtoull_impl(s, pe, b) _strtoui64_l(s, pe, b, ClassicLocale::Get())
+    #define __strtoll_impl(s, pe, b) _strtoi64_l(s, pe, b, ClassicLocale::Get())
+    #define __strtod_impl(s, pe) _strtod_l(s, pe, ClassicLocale::Get())
+    #define __strtof_impl(s, pe) _strtof_l(s, pe, ClassicLocale::Get())
+  #else
+    #define __strtoull_impl(s, pe, b) strtoull_l(s, pe, b, ClassicLocale::Get())
+    #define __strtoll_impl(s, pe, b) strtoll_l(s, pe, b, ClassicLocale::Get())
+    #define __strtod_impl(s, pe) strtod_l(s, pe, ClassicLocale::Get())
+    #define __strtof_impl(s, pe) strtof_l(s, pe, ClassicLocale::Get())
+  #endif
+#else
+  #define __strtod_impl(s, pe) strtod(s, pe)
+  #define __strtof_impl(s, pe) static_cast<float>(strtod(s, pe))
+  #ifdef _MSC_VER
+    #define __strtoull_impl(s, pe, b) _strtoui64(s, pe, b)
+    #define __strtoll_impl(s, pe, b) _strtoi64(s, pe, b)
+  #else
+    #define __strtoull_impl(s, pe, b) strtoull(s, pe, b)
+    #define __strtoll_impl(s, pe, b) strtoll(s, pe, b)
+  #endif
+#endif
+
+inline void strtoval_impl(int64_t *val, const char *str, char **endptr,
+                                 int base) {
+    *val = __strtoll_impl(str, endptr, base);
+}
+
+inline void strtoval_impl(uint64_t *val, const char *str, char **endptr,
+                                 int base) {
+  *val = __strtoull_impl(str, endptr, base);
+}
+
+inline void strtoval_impl(double *val, const char *str, char **endptr) {
+  *val = __strtod_impl(str, endptr);
+}
+
+// UBSAN: double to float is safe if numeric_limits<float>::is_iec559 is true.
+__supress_ubsan__("float-cast-overflow")
+inline void strtoval_impl(float *val, const char *str, char **endptr) {
+  *val = __strtof_impl(str, endptr);
+}
+#undef __strtoull_impl
+#undef __strtoll_impl
+#undef __strtod_impl
+#undef __strtof_impl
+// clang-format on
+
+// Adaptor for strtoull()/strtoll().
+// Flatbuffers accepts numbers with any count of leading zeros (-009 is -9),
+// while strtoll with base=0 interprets first leading zero as octal prefix.
+// In future, it is possible to add prefixed 0b0101.
+// 1) Checks errno code for overflow condition (out of range).
+// 2) If base <= 0, function try to detect base of number by prefix.
+//
+// Return value (like strtoull and strtoll, but reject partial result):
+// - If successful, an integer value corresponding to the str is returned.
+// - If full string conversion can't be performed, 0 is returned.
+// - If the converted value falls out of range of corresponding return type, a
+// range error occurs. In this case value MAX(T)/MIN(T) is returned.
+template<typename T>
+inline bool StringToIntegerImpl(T *val, const char *const str,
+                                const int base = 0,
+                                const bool check_errno = true) {
+  // T is int64_t or uint64_T
+  FLATBUFFERS_ASSERT(str);
+  if (base <= 0) {
+    auto s = str;
+    while (*s && !is_digit(*s)) s++;
+    if (s[0] == '0' && is_alpha_char(s[1], 'X'))
+      return StringToIntegerImpl(val, str, 16, check_errno);
+    // if a prefix not match, try base=10
+    return StringToIntegerImpl(val, str, 10, check_errno);
+  } else {
+    if (check_errno) errno = 0;  // clear thread-local errno
+    auto endptr = str;
+    strtoval_impl(val, str, const_cast<char **>(&endptr), base);
+    if ((*endptr != '\0') || (endptr == str)) {
+      *val = 0;      // erase partial result
+      return false;  // invalid string
+    }
+    // errno is out-of-range, return MAX/MIN
+    if (check_errno && errno) return false;
+    return true;
+  }
+}
+
+template<typename T>
+inline bool StringToFloatImpl(T *val, const char *const str) {
+  // Type T must be either float or double.
+  FLATBUFFERS_ASSERT(str && val);
+  auto end = str;
+  strtoval_impl(val, str, const_cast<char **>(&end));
+  auto done = (end != str) && (*end == '\0');
+  if (!done) *val = 0;  // erase partial result
+  return done;
+}
+
+// Convert a string to an instance of T.
+// Return value (matched with StringToInteger64Impl and strtod):
+// - If successful, a numeric value corresponding to the str is returned.
+// - If full string conversion can't be performed, 0 is returned.
+// - If the converted value falls out of range of corresponding return type, a
+// range error occurs. In this case value MAX(T)/MIN(T) is returned.
+template<typename T> inline bool StringToNumber(const char *s, T *val) {
+  FLATBUFFERS_ASSERT(s && val);
+  int64_t i64;
+  // The errno check isn't needed, will return MAX/MIN on overflow.
+  if (StringToIntegerImpl(&i64, s, 0, false)) {
+    const int64_t max = flatbuffers::numeric_limits<T>::max();
+    const int64_t min = flatbuffers::numeric_limits<T>::lowest();
+    if (i64 > max) {
+      *val = static_cast<T>(max);
+      return false;
+    }
+    if (i64 < min) {
+      // For unsigned types return max to distinguish from
+      // "no conversion can be performed" when 0 is returned.
+      *val = static_cast<T>(flatbuffers::is_unsigned<T>::value ? max : min);
+      return false;
+    }
+    *val = static_cast<T>(i64);
+    return true;
+  }
+  *val = 0;
+  return false;
+}
+
+template<> inline bool StringToNumber<int64_t>(const char *str, int64_t *val) {
+  return StringToIntegerImpl(val, str);
+}
+
+template<>
+inline bool StringToNumber<uint64_t>(const char *str, uint64_t *val) {
+  if (!StringToIntegerImpl(val, str)) return false;
+  // The strtoull accepts negative numbers:
+  // If the minus sign was part of the input sequence, the numeric value
+  // calculated from the sequence of digits is negated as if by unary minus
+  // in the result type, which applies unsigned integer wraparound rules.
+  // Fix this behaviour (except -0).
+  if (*val) {
+    auto s = str;
+    while (*s && !is_digit(*s)) s++;
+    s = (s > str) ? (s - 1) : s;  // step back to one symbol
+    if (*s == '-') {
+      // For unsigned types return the max to distinguish from
+      // "no conversion can be performed".
+      *val = flatbuffers::numeric_limits<uint64_t>::max();
+      return false;
+    }
+  }
+  return true;
+}
+
+template<> inline bool StringToNumber(const char *s, float *val) {
+  return StringToFloatImpl(val, s);
+}
+
+template<> inline bool StringToNumber(const char *s, double *val) {
+  return StringToFloatImpl(val, s);
+}
+
+inline int64_t StringToInt(const char *s, int base = 10) {
+  int64_t val;
+  return StringToIntegerImpl(&val, s, base) ? val : 0;
+}
+
+inline uint64_t StringToUInt(const char *s, int base = 10) {
+  uint64_t val;
+  return StringToIntegerImpl(&val, s, base) ? val : 0;
+}
+
+typedef bool (*LoadFileFunction)(const char *filename, bool binary,
+                                 std::string *dest);
+typedef bool (*FileExistsFunction)(const char *filename);
+
+LoadFileFunction SetLoadFileFunction(LoadFileFunction load_file_function);
+
+FileExistsFunction SetFileExistsFunction(
+    FileExistsFunction file_exists_function);
+
+// Check if file "name" exists.
+bool FileExists(const char *name);
+
+// Check if "name" exists and it is also a directory.
+bool DirExists(const char *name);
+
+// Load file "name" into "buf" returning true if successful
+// false otherwise.  If "binary" is false data is read
+// using ifstream's text mode, otherwise data is read with
+// no transcoding.
+bool LoadFile(const char *name, bool binary, std::string *buf);
+
+// Save data "buf" of length "len" bytes into a file
+// "name" returning true if successful, false otherwise.
+// If "binary" is false data is written using ifstream's
+// text mode, otherwise data is written with no
+// transcoding.
+bool SaveFile(const char *name, const char *buf, size_t len, bool binary);
+
+// Save data "buf" into file "name" returning true if
+// successful, false otherwise.  If "binary" is false
+// data is written using ifstream's text mode, otherwise
+// data is written with no transcoding.
+inline bool SaveFile(const char *name, const std::string &buf, bool binary) {
+  return SaveFile(name, buf.c_str(), buf.size(), binary);
+}
+
+// Functionality for minimalistic portable path handling.
+
+// The functions below behave correctly regardless of whether posix ('/') or
+// Windows ('/' or '\\') separators are used.
+
+// Any new separators inserted are always posix.
+FLATBUFFERS_CONSTEXPR char kPathSeparator = '/';
+
+// Returns the path with the extension, if any, removed.
+std::string StripExtension(const std::string &filepath);
+
+// Returns the extension, if any.
+std::string GetExtension(const std::string &filepath);
+
+// Return the last component of the path, after the last separator.
+std::string StripPath(const std::string &filepath);
+
+// Strip the last component of the path + separator.
+std::string StripFileName(const std::string &filepath);
+
+// Concatenates a path with a filename, regardless of wether the path
+// ends in a separator or not.
+std::string ConCatPathFileName(const std::string &path,
+                               const std::string &filename);
+
+// Replaces any '\\' separators with '/'
+std::string PosixPath(const char *path);
+
+// This function ensure a directory exists, by recursively
+// creating dirs for any parts of the path that don't exist yet.
+void EnsureDirExists(const std::string &filepath);
+
+// Obtains the absolute path from any other path.
+// Returns the input path if the absolute path couldn't be resolved.
+std::string AbsolutePath(const std::string &filepath);
+
+// To and from UTF-8 unicode conversion functions
+
+// Convert a unicode code point into a UTF-8 representation by appending it
+// to a string. Returns the number of bytes generated.
+inline int ToUTF8(uint32_t ucc, std::string *out) {
+  FLATBUFFERS_ASSERT(!(ucc & 0x80000000));  // Top bit can't be set.
+  // 6 possible encodings: http://en.wikipedia.org/wiki/UTF-8
+  for (int i = 0; i < 6; i++) {
+    // Max bits this encoding can represent.
+    uint32_t max_bits = 6 + i * 5 + static_cast<int>(!i);
+    if (ucc < (1u << max_bits)) {  // does it fit?
+      // Remaining bits not encoded in the first byte, store 6 bits each
+      uint32_t remain_bits = i * 6;
+      // Store first byte:
+      (*out) += static_cast<char>((0xFE << (max_bits - remain_bits)) |
+                                  (ucc >> remain_bits));
+      // Store remaining bytes:
+      for (int j = i - 1; j >= 0; j--) {
+        (*out) += static_cast<char>(((ucc >> (j * 6)) & 0x3F) | 0x80);
+      }
+      return i + 1;  // Return the number of bytes added.
+    }
+  }
+  FLATBUFFERS_ASSERT(0);  // Impossible to arrive here.
+  return -1;
+}
+
+// Converts whatever prefix of the incoming string corresponds to a valid
+// UTF-8 sequence into a unicode code. The incoming pointer will have been
+// advanced past all bytes parsed.
+// returns -1 upon corrupt UTF-8 encoding (ignore the incoming pointer in
+// this case).
+inline int FromUTF8(const char **in) {
+  int len = 0;
+  // Count leading 1 bits.
+  for (int mask = 0x80; mask >= 0x04; mask >>= 1) {
+    if (**in & mask) {
+      len++;
+    } else {
+      break;
+    }
+  }
+  if ((static_cast<unsigned char>(**in) << len) & 0x80)
+    return -1;  // Bit after leading 1's must be 0.
+  if (!len) return *(*in)++;
+  // UTF-8 encoded values with a length are between 2 and 4 bytes.
+  if (len < 2 || len > 4) { return -1; }
+  // Grab initial bits of the code.
+  int ucc = *(*in)++ & ((1 << (7 - len)) - 1);
+  for (int i = 0; i < len - 1; i++) {
+    if ((**in & 0xC0) != 0x80) return -1;  // Upper bits must 1 0.
+    ucc <<= 6;
+    ucc |= *(*in)++ & 0x3F;  // Grab 6 more bits of the code.
+  }
+  // UTF-8 cannot encode values between 0xD800 and 0xDFFF (reserved for
+  // UTF-16 surrogate pairs).
+  if (ucc >= 0xD800 && ucc <= 0xDFFF) { return -1; }
+  // UTF-8 must represent code points in their shortest possible encoding.
+  switch (len) {
+    case 2:
+      // Two bytes of UTF-8 can represent code points from U+0080 to U+07FF.
+      if (ucc < 0x0080 || ucc > 0x07FF) { return -1; }
+      break;
+    case 3:
+      // Three bytes of UTF-8 can represent code points from U+0800 to U+FFFF.
+      if (ucc < 0x0800 || ucc > 0xFFFF) { return -1; }
+      break;
+    case 4:
+      // Four bytes of UTF-8 can represent code points from U+10000 to U+10FFFF.
+      if (ucc < 0x10000 || ucc > 0x10FFFF) { return -1; }
+      break;
+  }
+  return ucc;
+}
+
+#ifndef FLATBUFFERS_PREFER_PRINTF
+// Wraps a string to a maximum length, inserting new lines where necessary. Any
+// existing whitespace will be collapsed down to a single space. A prefix or
+// suffix can be provided, which will be inserted before or after a wrapped
+// line, respectively.
+inline std::string WordWrap(const std::string in, size_t max_length,
+                            const std::string wrapped_line_prefix,
+                            const std::string wrapped_line_suffix) {
+  std::istringstream in_stream(in);
+  std::string wrapped, line, word;
+
+  in_stream >> word;
+  line = word;
+
+  while (in_stream >> word) {
+    if ((line.length() + 1 + word.length() + wrapped_line_suffix.length()) <
+        max_length) {
+      line += " " + word;
+    } else {
+      wrapped += line + wrapped_line_suffix + "\n";
+      line = wrapped_line_prefix + word;
+    }
+  }
+  wrapped += line;
+
+  return wrapped;
+}
+#endif  // !FLATBUFFERS_PREFER_PRINTF
+
+inline bool EscapeString(const char *s, size_t length, std::string *_text,
+                         bool allow_non_utf8, bool natural_utf8) {
+  std::string &text = *_text;
+  text += "\"";
+  for (uoffset_t i = 0; i < length; i++) {
+    char c = s[i];
+    switch (c) {
+      case '\n': text += "\\n"; break;
+      case '\t': text += "\\t"; break;
+      case '\r': text += "\\r"; break;
+      case '\b': text += "\\b"; break;
+      case '\f': text += "\\f"; break;
+      case '\"': text += "\\\""; break;
+      case '\\': text += "\\\\"; break;
+      default:
+        if (c >= ' ' && c <= '~') {
+          text += c;
+        } else {
+          // Not printable ASCII data. Let's see if it's valid UTF-8 first:
+          const char *utf8 = s + i;
+          int ucc = FromUTF8(&utf8);
+          if (ucc < 0) {
+            if (allow_non_utf8) {
+              text += "\\x";
+              text += IntToStringHex(static_cast<uint8_t>(c), 2);
+            } else {
+              // There are two cases here:
+              //
+              // 1) We reached here by parsing an IDL file. In that case,
+              // we previously checked for non-UTF-8, so we shouldn't reach
+              // here.
+              //
+              // 2) We reached here by someone calling GenerateText()
+              // on a previously-serialized flatbuffer. The data might have
+              // non-UTF-8 Strings, or might be corrupt.
+              //
+              // In both cases, we have to give up and inform the caller
+              // they have no JSON.
+              return false;
+            }
+          } else {
+            if (natural_utf8) {
+              // utf8 points to past all utf-8 bytes parsed
+              text.append(s + i, static_cast<size_t>(utf8 - s - i));
+            } else if (ucc <= 0xFFFF) {
+              // Parses as Unicode within JSON's \uXXXX range, so use that.
+              text += "\\u";
+              text += IntToStringHex(ucc, 4);
+            } else if (ucc <= 0x10FFFF) {
+              // Encode Unicode SMP values to a surrogate pair using two \u
+              // escapes.
+              uint32_t base = ucc - 0x10000;
+              auto high_surrogate = (base >> 10) + 0xD800;
+              auto low_surrogate = (base & 0x03FF) + 0xDC00;
+              text += "\\u";
+              text += IntToStringHex(high_surrogate, 4);
+              text += "\\u";
+              text += IntToStringHex(low_surrogate, 4);
+            }
+            // Skip past characters recognized.
+            i = static_cast<uoffset_t>(utf8 - s - 1);
+          }
+        }
+        break;
+    }
+  }
+  text += "\"";
+  return true;
+}
+
+// Remove paired quotes in a string: "text"|'text' -> text.
+std::string RemoveStringQuotes(const std::string &s);
+
+// Change th global C-locale to locale with name <locale_name>.
+// Returns an actual locale name in <_value>, useful if locale_name is "" or
+// null.
+bool SetGlobalTestLocale(const char *locale_name,
+                         std::string *_value = nullptr);
+
+// Read (or test) a value of environment variable.
+bool ReadEnvironmentVariable(const char *var_name,
+                             std::string *_value = nullptr);
+
+}  // namespace MindSpore.flatbuffers
+
+#endif  // FLATBUFFERS_UTIL_H_

model_zoo/official/lite/app/src/main/cpp/include/MindSpore/ir/dtype/type_id.h

@@ -0,0 +1,87 @@
+/**
+ * This is the C++ adaptation and derivative work of Myia (https://github.com/mila-iqia/myia/).
+ *
+ * Copyright 2019-2020 Huawei Technologies Co., Ltd
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef MINDSPORE_CORE_IR_DTYPE_TYPE_ID_H_
+#define MINDSPORE_CORE_IR_DTYPE_TYPE_ID_H_
+
+#include <unordered_map>
+#include <string>
+
+namespace mindspore {
+//
+// Supported meta type
+//
+enum TypeId : int {
+  kTypeUnknown = 0,
+  kMetaTypeBegin = kTypeUnknown,
+  kMetaTypeType,  // Type
+  kMetaTypeAnything,
+  kMetaTypeObject,
+  kMetaTypeTypeType,  // TypeType
+  kMetaTypeProblem,
+  kMetaTypeExternal,
+  kMetaTypeNone,
+  kMetaTypeNull,
+  kMetaTypeEllipsis,
+  kMetaTypeEnd,
+  //
+  // Object types
+  //
+  kObjectTypeBegin = kMetaTypeEnd,
+  kObjectTypeNumber,
+  kObjectTypeString,
+  kObjectTypeList,
+  kObjectTypeTuple,
+  kObjectTypeSlice,
+  kObjectTypeKeyword,
+  kObjectTypeTensorType,
+  kObjectTypeIndexedSlicesType,
+  kObjectTypeSparseTensorType,
+  kObjectTypeUndeterminedType,
+  kObjectTypeClass,
+  kObjectTypeDictionary,
+  kObjectTypeFunction,
+  kObjectTypeJTagged,
+  kObjectTypeSymbolicKeyType,
+  kObjectTypeEnvType,
+  kObjectTypeRefKey,
+  kObjectTypeRef,
+  kObjectTypeEnd,
+  //
+  // Number Types
+  //
+  kNumberTypeBegin = kObjectTypeEnd,
+  kNumberTypeBool,
+  kNumberTypeInt,
+  kNumberTypeInt8,
+  kNumberTypeInt16,
+  kNumberTypeInt32,
+  kNumberTypeInt64,
+  kNumberTypeUInt,
+  kNumberTypeUInt8,
+  kNumberTypeUInt16,
+  kNumberTypeUInt32,
+  kNumberTypeUInt64,
+  kNumberTypeFloat,
+  kNumberTypeFloat16,
+  kNumberTypeFloat32,
+  kNumberTypeFloat64,
+  kNumberTypeEnd
+};
+}  // namespace mindspore
+#endif  // MINDSPORE_CORE_IR_DTYPE_TYPE_ID_H_

model_zoo/official/lite/app/src/main/cpp/include/MindSpore/lite_session.h

@@ -0,0 +1,127 @@
+/**
+ * Copyright 2020 Huawei Technologies Co., Ltd
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef MINDSPORE_LITE_INCLUDE_LITE_SESSION_H
+#define MINDSPORE_LITE_INCLUDE_LITE_SESSION_H
+
+#include <memory>
+#include <vector>
+#include <string>
+#include <unordered_map>
+#include "ms_tensor.h"
+#include "model.h"
+#include "context.h"
+
+namespace mindspore {
+namespace session {
+/// \brief CallBackParam defined input arguments for callBack function.
+struct CallBackParam {
+  std::string name_callback_param; /**< node name argument */
+  std::string type_callback_param; /**< node type argument */
+};
+
+/// \brief KernelCallBack defined the function pointer for callBack.
+using KernelCallBack = std::function<bool(std::vector<tensor::MSTensor *> inputs,
+                                          std::vector<tensor::MSTensor *> outputs, const CallBackParam &opInfo)>;
+
+/// \brief LiteSession defined session in MindSpore Lite for compiling Model and forwarding model.
+class MS_API LiteSession {
+ public:
+  /// \brief Static method to create a LiteSession pointer.
+  ///
+  /// \param[in] context Define the context of session to be created.
+  ///
+  /// \return Pointer of MindSpore Lite LiteSession.
+  static LiteSession *CreateSession(lite::Context *context);
+
+  /// \brief Destructor of MindSpore Lite LiteSession.
+  virtual ~LiteSession() = default;
+
+  /// \brief Attempt to bind or unbind threads in the thread pool to or from the specified cpu core.
+  ///
+  /// \param[in] if_bind Define whether to bind or unbind threads.
+  virtual void BindThread(bool if_bind) = 0;
+
+  /// \brief Compile MindSpore Lite model.
+  ///
+  /// \note CompileGraph should be called before RunGraph.
+  ///
+  /// \param[in] model Define the model to be compiled.
+  ///
+  /// \return STATUS as an error code of compiling graph, STATUS is defined in errorcode.h.
+  virtual int CompileGraph(lite::Model *model) = 0;
+
+  /// \brief Get input MindSpore Lite MSTensors of model.
+  ///
+  /// \return The vector of MindSpore Lite MSTensor.
+  virtual std::vector<tensor::MSTensor *> GetInputs() const = 0;
+
+  /// \brief Get input MindSpore Lite MSTensors of model by node name.
+  ///
+  /// \param[in] node_name Define node name.
+  ///
+  /// \return The vector of MindSpore Lite MSTensor.
+  virtual std::vector<tensor::MSTensor *> GetInputsByName(const std::string &node_name) const = 0;
+
+  /// \brief Run session with callback.
+  ///
+  /// \param[in] before Define a call_back_function to be called before running each node.
+  /// \param[in] after Define a call_back_function called after running each node.
+  ///
+  /// \note RunGraph should be called after CompileGraph.
+  ///
+  /// \return STATUS as an error code of running graph, STATUS is defined in errorcode.h.
+  virtual int RunGraph(const KernelCallBack &before = nullptr, const KernelCallBack &after = nullptr) = 0;
+
+  /// \brief Get output MindSpore Lite MSTensors of model mapped by node name.
+  ///
+  /// \return The map of output node name and MindSpore Lite MSTensor.
+  virtual std::unordered_map<std::string, std::vector<mindspore::tensor::MSTensor *>> GetOutputMapByNode() const = 0;
+
+  /// \brief Get output MindSpore Lite MSTensors of model by node name.
+  ///
+  /// \param[in] node_name Define node name.
+  ///
+  /// \return The vector of MindSpore Lite MSTensor.
+  virtual std::vector<tensor::MSTensor *> GetOutputsByNodeName(const std::string &node_name) const = 0;
+
+  /// \brief Get output MindSpore Lite MSTensors of model mapped by tensor name.
+  ///
+  /// \return The map of output tensor name and MindSpore Lite MSTensor.
+  virtual std::unordered_map<std::string, mindspore::tensor::MSTensor *> GetOutputMapByTensor() const = 0;
+
+  /// \brief Get name of output tensors of model compiled by this session.
+  ///
+  /// \return The vector of string as output tensor names in order.
+  virtual std::vector<std::string> GetOutputTensorNames() const = 0;
+
+  /// \brief Get output MindSpore Lite MSTensors of model by tensor name.
+  ///
+  /// \param[in] tensor_name Define tensor name.
+  ///
+  /// \return Pointer of MindSpore Lite MSTensor.
+  virtual mindspore::tensor::MSTensor *GetOutputByTensorName(const std::string &tensor_name) const = 0;
+
+  /// \brief Resize inputs shape.
+  ///
+  /// \param[in] inputs Define the new inputs shape.
+  ///
+  /// \return STATUS as an error code of resize inputs, STATUS is defined in errorcode.h.
+  virtual int Resize(const std::vector<tensor::MSTensor *> &inputs) = 0;
+};
+}  // namespace session
+}  // namespace mindspore
+#endif  // MINDSPORE_LITE_INCLUDE_LITE_SESSION_H

model_zoo/official/lite/app/src/main/cpp/include/MindSpore/model.h

@@ -0,0 +1,110 @@
+/**
+ * Copyright 2020 Huawei Technologies Co., Ltd
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef MINDSPORE_LITE_INCLUDE_MODEL_H
+#define MINDSPORE_LITE_INCLUDE_MODEL_H
+
+#include <string>
+#include <vector>
+#include <memory>
+#include "schema/model_generated.h"
+
+namespace mindspore {
+#define MS_API __attribute__((visibility("default")))
+
+namespace lite {
+/// \brief ModelImpl defined the implement class of Model in MindSpore Lite.
+///
+/// \note List public class and interface for reference.
+class ModelImpl;
+
+/// \brief Primitive defined as prototype of operator.
+///
+/// \note List public class and interface for reference.
+class PrimitiveC;
+
+/// \brief Model defined model in MindSpore Lite for managing graph.
+class MS_API Model {
+ public:
+  /// \brief Static method to create a Model pointer.
+  ///
+  /// \param[in] model_buf Define the buffer read from a model file.
+  /// \param[in] size Define bytes number of model buffer.
+  ///
+  /// \return Pointer of MindSpore Lite Model.
+  static Model *Import(const char *model_buf, size_t size);
+
+  /// \brief Constructor of MindSpore Lite Model using default value for parameters.
+  ///
+  /// \return Instance of MindSpore Lite Model.
+  Model() = default;
+
+  /// \brief Destructor of MindSpore Lite Model.
+  virtual ~Model();
+
+  /// \brief Get MindSpore Lite Primitive by name.
+  ///
+  /// \param[in] name Define name of primitive to be returned.
+  ///
+  /// \return the pointer of MindSpore Lite Primitive.
+  PrimitiveC *GetOp(const std::string &name) const;
+
+  /// \brief Get graph defined in flatbuffers.
+  ///
+  /// \return the pointer of graph defined in flatbuffers.
+  const schema::MetaGraph *GetMetaGraph() const;
+
+  /// \brief Free MetaGraph in MindSpore Lite Model.
+  void FreeMetaGraph();
+
+ protected:
+  ModelImpl *model_impl_ = nullptr;
+};
+
+/// \brief ModelBuilder defined by MindSpore Lite.
+class MS_API ModelBuilder {
+ public:
+  /// \brief OutEdge defined by MindSpore Lite.
+  struct OutEdge {
+    std::string nodeId;  /**< ID of a node linked by this edge */
+    size_t outEdgeIndex; /**< Index of this edge */
+  };
+
+  /// \brief Constructor of MindSpore Lite Model using default value for parameters.
+  ///
+  /// \return Instance of MindSpore Lite ModelBuilder.
+  ModelBuilder() = default;
+
+  /// \brief Destructor of MindSpore Lite ModelBuilder.
+  virtual ~ModelBuilder() = default;
+
+  /// \brief Add primitive into model builder for model building.
+  ///
+  /// \param[in] op Define the primitive to be added.
+  /// \param[in] inputs Define input edge of primitive to be added.
+  ///
+  /// \return ID of the added primitive.
+  virtual std::string AddOp(const PrimitiveC &op, const std::vector<OutEdge> &inputs) = 0;
+
+  /// \brief Finish constructing the model.
+  ///
+  /// \return the pointer of MindSpore Lite Model.
+  virtual Model *Construct();
+};
+}  // namespace lite
+}  // namespace mindspore
+
+#endif  // MINDSPORE_LITE_INCLUDE_MODEL_H

model_zoo/official/lite/app/src/main/cpp/include/MindSpore/ms_tensor.h

@@ -0,0 +1,108 @@
+/**
+ * Copyright 2020 Huawei Technologies Co., Ltd
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef MINDSPORE_LITE_INCLUDE_MS_TENSOR_H_
+#define MINDSPORE_LITE_INCLUDE_MS_TENSOR_H_
+
+#include <utility>
+#include <vector>
+#include <memory>
+#include "ir/dtype/type_id.h"
+
+namespace mindspore {
+#define MS_API __attribute__((visibility("default")))
+namespace tensor {
+/// \brief MSTensor defined tensor in MindSpore Lite.
+class MS_API MSTensor {
+ public:
+  /// \brief Constructor of MindSpore Lite MSTensor.
+  ///
+  /// \return Instance of MindSpore Lite MSTensor.
+  MSTensor() = default;
+
+  /// \brief Static method to create a MSTensor pointer.
+  ///
+  /// \param[in] data_type Define data type of tensor to be created.
+  /// \param[in] shape Define Shape of tensor to be created.
+  ///
+  /// \note TypeId is defined in mindspore/mindspore/core/ir/dtype/type_id.h. Only number types in TypeId enum are
+  /// suitable for MSTensor.
+  ///
+  /// \return the pointer of MSTensor.
+  static MSTensor *CreateTensor(TypeId data_type, const std::vector<int> &shape);
+
+  /// \brief Destructor of MindSpore Lite Model.
+  virtual ~MSTensor() = default;
+
+  /// \brief Get data type of the MindSpore Lite MSTensor.
+  ///
+  /// \note TypeId is defined in mindspore/mindspore/core/ir/dtype/type_id.h. Only number types in TypeId enum are
+  /// suitable for MSTensor.
+  ///
+  /// \return MindSpore Lite TypeId of the MindSpore Lite MSTensor.
+  virtual TypeId data_type() const = 0;
+
+  /// \brief Set data type for the MindSpore Lite MSTensor.
+  ///
+  /// \param[in] data_type Define MindSpore Lite TypeId to be set in the MindSpore Lite MSTensor.
+  ///
+  /// \return MindSpore Lite TypeId of the MindSpore Lite MSTensor after set.
+  virtual TypeId set_data_type(TypeId data_type) = 0;
+
+  /// \brief Get shape of the MindSpore Lite MSTensor.
+  ///
+  /// \return A vector of int as the shape of the MindSpore Lite MSTensor.
+  virtual std::vector<int> shape() const = 0;
+
+  /// \brief Set shape for the MindSpore Lite MSTensor.
+  ///
+  /// \param[in] shape Define a vector of int as shape to be set into the MindSpore Lite MSTensor.
+  ///
+  /// \return size of shape of the MindSpore Lite MSTensor after set.
+  virtual size_t set_shape(const std::vector<int> &shape) = 0;
+
+  /// \brief Get size of the dimension of the MindSpore Lite MSTensor index by the parameter index.
+  ///
+  /// \param[in] index Define index of dimension returned.
+  ///
+  /// \return Size of dimension of the MindSpore Lite MSTensor.
+  virtual int DimensionSize(size_t index) const = 0;
+
+  /// \brief Get number of element in MSTensor.
+  ///
+  /// \return Number of element in MSTensor.
+  virtual int ElementsNum() const = 0;
+
+  /// \brief Get hash of the MindSpore Lite MSTensor.
+  ///
+  /// \return Hash of the MindSpore Lite MSTensor.
+  virtual std::size_t hash() const = 0;
+
+  /// \brief Get byte size of data in MSTensor.
+  ///
+  /// \return Byte size of data in MSTensor.
+  virtual size_t Size() const = 0;
+
+  /// \brief Get the pointer of data in MSTensor.
+  ///
+  /// \note The data pointer can be used to both write and read data in MSTensor.
+  ///
+  /// \return the pointer points to data in MSTensor.
+  virtual void *MutableData() const = 0;
+};
+}  // namespace tensor
+}  // namespace mindspore
+#endif  // MINDSPORE_LITE_INCLUDE_MS_TENSOR_H_

model_zoo/official/lite/app/src/main/cpp/include/MindSpore/thread_pool_config.h

@@ -0,0 +1,35 @@
+/**
+ * Copyright 2020 Huawei Technologies Co., Ltd
+ *
+ * Licensed under the Apache License, Version 2.0 (the "License");
+ * you may not use this file except in compliance with the License.
+ * You may obtain a copy of the License at
+ *
+ * http://www.apache.org/licenses/LICENSE-2.0
+ *
+ * Unless required by applicable law or agreed to in writing, software
+ * distributed under the License is distributed on an "AS IS" BASIS,
+ * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+ * See the License for the specific language governing permissions and
+ * limitations under the License.
+ */
+
+#ifndef MINDSPORE_LITE_INCLUDE_THREAD_POOL_CONFIG_H_
+#define MINDSPORE_LITE_INCLUDE_THREAD_POOL_CONFIG_H_
+
+/// \brief CpuBindMode defined for holding bind cpu strategy argument.
+typedef enum Mode {
+  MID_CPU = -1,   /**< bind middle cpu first */
+  HIGHER_CPU = 1, /**< bind higher cpu first */
+  NO_BIND = 0     /**< no bind */
+} CpuBindMode;
+
+/// \brief ThreadPoolId defined for specifying which thread pool to use.
+typedef enum Id {
+  THREAD_POOL_DEFAULT = 0, /**< default thread pool id */
+  THREAD_POOL_SECOND = 1,  /**< the second thread pool id */
+  THREAD_POOL_THIRD = 2,   /**< the third thread pool id */
+  THREAD_POOL_FOURTH = 3   /**< the fourth thread pool id */
+} ThreadPoolId;
+
+#endif  // LITE_MINDSPORE_LITE_INCLUDE_THREAD_POOL_CONFIG_H_

model_zoo/official/lite/app/src/main/cpp/include/opencv2/calib3d/calib3d.hpp

@@ -0,0 +1,48 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifdef __OPENCV_BUILD
+#error this is a compatibility header which should not be used inside the OpenCV library
+#endif
+
+#include "opencv2/calib3d.hpp"

model_zoo/official/lite/app/src/main/cpp/include/opencv2/calib3d/calib3d_c.h

@@ -0,0 +1,150 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CALIB3D_C_H
+#define OPENCV_CALIB3D_C_H
+
+#include "opencv2/core/types_c.h"
+
+#ifdef __cplusplus
+extern "C" {
+#endif
+
+/* Calculates fundamental matrix given a set of corresponding points */
+#define CV_FM_7POINT 1
+#define CV_FM_8POINT 2
+
+#define CV_LMEDS 4
+#define CV_RANSAC 8
+
+#define CV_FM_LMEDS_ONLY  CV_LMEDS
+#define CV_FM_RANSAC_ONLY CV_RANSAC
+#define CV_FM_LMEDS CV_LMEDS
+#define CV_FM_RANSAC CV_RANSAC
+
+enum
+{
+    CV_ITERATIVE = 0,
+    CV_EPNP = 1, // F.Moreno-Noguer, V.Lepetit and P.Fua "EPnP: Efficient Perspective-n-Point Camera Pose Estimation"
+    CV_P3P = 2, // X.S. Gao, X.-R. Hou, J. Tang, H.-F. Chang; "Complete Solution Classification for the Perspective-Three-Point Problem"
+    CV_DLS = 3 // Joel A. Hesch and Stergios I. Roumeliotis. "A Direct Least-Squares (DLS) Method for PnP"
+};
+
+#define CV_CALIB_CB_ADAPTIVE_THRESH  1
+#define CV_CALIB_CB_NORMALIZE_IMAGE  2
+#define CV_CALIB_CB_FILTER_QUADS     4
+#define CV_CALIB_CB_FAST_CHECK       8
+
+#define CV_CALIB_USE_INTRINSIC_GUESS  1
+#define CV_CALIB_FIX_ASPECT_RATIO     2
+#define CV_CALIB_FIX_PRINCIPAL_POINT  4
+#define CV_CALIB_ZERO_TANGENT_DIST    8
+#define CV_CALIB_FIX_FOCAL_LENGTH 16
+#define CV_CALIB_FIX_K1  32
+#define CV_CALIB_FIX_K2  64
+#define CV_CALIB_FIX_K3  128
+#define CV_CALIB_FIX_K4  2048
+#define CV_CALIB_FIX_K5  4096
+#define CV_CALIB_FIX_K6  8192
+#define CV_CALIB_RATIONAL_MODEL 16384
+#define CV_CALIB_THIN_PRISM_MODEL 32768
+#define CV_CALIB_FIX_S1_S2_S3_S4  65536
+#define CV_CALIB_TILTED_MODEL  262144
+#define CV_CALIB_FIX_TAUX_TAUY  524288
+#define CV_CALIB_FIX_TANGENT_DIST 2097152
+
+#define CV_CALIB_NINTRINSIC 18
+
+#define CV_CALIB_FIX_INTRINSIC  256
+#define CV_CALIB_SAME_FOCAL_LENGTH 512
+
+#define CV_CALIB_ZERO_DISPARITY 1024
+
+/* stereo correspondence parameters and functions */
+#define CV_STEREO_BM_NORMALIZED_RESPONSE  0
+#define CV_STEREO_BM_XSOBEL               1
+
+#ifdef __cplusplus
+} // extern "C"
+
+//////////////////////////////////////////////////////////////////////////////////////////
+class CV_EXPORTS CvLevMarq
+{
+public:
+    CvLevMarq();
+    CvLevMarq( int nparams, int nerrs, CvTermCriteria criteria=
+              cvTermCriteria(CV_TERMCRIT_EPS+CV_TERMCRIT_ITER,30,DBL_EPSILON),
+              bool completeSymmFlag=false );
+    ~CvLevMarq();
+    void init( int nparams, int nerrs, CvTermCriteria criteria=
+              cvTermCriteria(CV_TERMCRIT_EPS+CV_TERMCRIT_ITER,30,DBL_EPSILON),
+              bool completeSymmFlag=false );
+    bool update( const CvMat*& param, CvMat*& J, CvMat*& err );
+    bool updateAlt( const CvMat*& param, CvMat*& JtJ, CvMat*& JtErr, double*& errNorm );
+
+    void clear();
+    void step();
+    enum { DONE=0, STARTED=1, CALC_J=2, CHECK_ERR=3 };
+
+    cv::Ptr<CvMat> mask;
+    cv::Ptr<CvMat> prevParam;
+    cv::Ptr<CvMat> param;
+    cv::Ptr<CvMat> J;
+    cv::Ptr<CvMat> err;
+    cv::Ptr<CvMat> JtJ;
+    cv::Ptr<CvMat> JtJN;
+    cv::Ptr<CvMat> JtErr;
+    cv::Ptr<CvMat> JtJV;
+    cv::Ptr<CvMat> JtJW;
+    double prevErrNorm, errNorm;
+    int lambdaLg10;
+    CvTermCriteria criteria;
+    int state;
+    int iters;
+    bool completeSymmFlag;
+    int solveMethod;
+};
+
+#endif
+
+#endif /* OPENCV_CALIB3D_C_H */

model_zoo/official/lite/app/src/main/cpp/include/opencv2/core/affine.hpp

@@ -0,0 +1,678 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CORE_AFFINE3_HPP
+#define OPENCV_CORE_AFFINE3_HPP
+
+#ifdef __cplusplus
+
+#include <opencv2/core.hpp>
+
+namespace cv
+{
+
+//! @addtogroup core
+//! @{
+
+    /** @brief Affine transform
+     *
+     * It represents a 4x4 homogeneous transformation matrix \f$T\f$
+     *
+     *  \f[T =
+     *  \begin{bmatrix}
+     *  R & t\\
+     *  0 & 1\\
+     *  \end{bmatrix}
+     *  \f]
+     *
+     *  where \f$R\f$ is a 3x3 rotation matrix and \f$t\f$ is a 3x1 translation vector.
+     *
+     *  You can specify \f$R\f$ either by a 3x3 rotation matrix or by a 3x1 rotation vector,
+     *  which is converted to a 3x3 rotation matrix by the Rodrigues formula.
+     *
+     *  To construct a matrix \f$T\f$ representing first rotation around the axis \f$r\f$ with rotation
+     *  angle \f$|r|\f$ in radian (right hand rule) and then translation by the vector \f$t\f$, you can use
+     *
+     *  @code
+     *  cv::Vec3f r, t;
+     *  cv::Affine3f T(r, t);
+     *  @endcode
+     *
+     *  If you already have the rotation matrix \f$R\f$, then you can use
+     *
+     *  @code
+     *  cv::Matx33f R;
+     *  cv::Affine3f T(R, t);
+     *  @endcode
+     *
+     *  To extract the rotation matrix \f$R\f$ from \f$T\f$, use
+     *
+     *  @code
+     *  cv::Matx33f R = T.rotation();
+     *  @endcode
+     *
+     *  To extract the translation vector \f$t\f$ from \f$T\f$, use
+     *
+     *  @code
+     *  cv::Vec3f t = T.translation();
+     *  @endcode
+     *
+     *  To extract the rotation vector \f$r\f$ from \f$T\f$, use
+     *
+     *  @code
+     *  cv::Vec3f r = T.rvec();
+     *  @endcode
+     *
+     *  Note that since the mapping from rotation vectors to rotation matrices
+     *  is many to one. The returned rotation vector is not necessarily the one
+     *  you used before to set the matrix.
+     *
+     *  If you have two transformations \f$T = T_1 * T_2\f$, use
+     *
+     *  @code
+     *  cv::Affine3f T, T1, T2;
+     *  T = T2.concatenate(T1);
+     *  @endcode
+     *
+     *  To get the inverse transform of \f$T\f$, use
+     *
+     *  @code
+     *  cv::Affine3f T, T_inv;
+     *  T_inv = T.inv();
+     *  @endcode
+     *
+     */
+    template<typename T>
+    class Affine3
+    {
+    public:
+        typedef T float_type;
+        typedef Matx<float_type, 3, 3> Mat3;
+        typedef Matx<float_type, 4, 4> Mat4;
+        typedef Vec<float_type, 3> Vec3;
+
+       //! Default constructor. It represents a 4x4 identity matrix.
+        Affine3();
+
+        //! Augmented affine matrix
+        Affine3(const Mat4& affine);
+
+        /**
+         *  The resulting 4x4 matrix is
+         *
+         *  \f[
+         *  \begin{bmatrix}
+         *  R & t\\
+         *  0 & 1\\
+         *  \end{bmatrix}
+         *  \f]
+         *
+         * @param R 3x3 rotation matrix.
+         * @param t 3x1 translation vector.
+         */
+        Affine3(const Mat3& R, const Vec3& t = Vec3::all(0));
+
+        /**
+         * Rodrigues vector.
+         *
+         * The last row of the current matrix is set to [0,0,0,1].
+         *
+         * @param rvec 3x1 rotation vector. Its direction indicates the rotation axis and its length
+         *             indicates the rotation angle in radian (using right hand rule).
+         * @param t 3x1 translation vector.
+         */
+        Affine3(const Vec3& rvec, const Vec3& t = Vec3::all(0));
+
+        /**
+         * Combines all constructors above. Supports 4x4, 3x4, 3x3, 1x3, 3x1 sizes of data matrix.
+         *
+         * The last row of the current matrix is set to [0,0,0,1] when data is not 4x4.
+         *
+         * @param data 1-channel matrix.
+         *             when it is 4x4, it is copied to the current matrix and t is not used.
+         *             When it is 3x4, it is copied to the upper part 3x4 of the current matrix and t is not used.
+         *             When it is 3x3, it is copied to the upper left 3x3 part of the current matrix.
+         *             When it is 3x1 or 1x3, it is treated as a rotation vector and the Rodrigues formula is used
+         *                             to compute a 3x3 rotation matrix.
+         * @param t 3x1 translation vector. It is used only when data is neither 4x4 nor 3x4.
+         */
+        explicit Affine3(const Mat& data, const Vec3& t = Vec3::all(0));
+
+        //! From 16-element array
+        explicit Affine3(const float_type* vals);
+
+        //! Create an 4x4 identity transform
+        static Affine3 Identity();
+
+        /**
+         * Rotation matrix.
+         *
+         * Copy the rotation matrix to the upper left 3x3 part of the current matrix.
+         * The remaining elements of the current matrix are not changed.
+         *
+         * @param R 3x3 rotation matrix.
+         *
+         */
+        void rotation(const Mat3& R);
+
+        /**
+         * Rodrigues vector.
+         *
+         * It sets the upper left 3x3 part of the matrix. The remaining part is unaffected.
+         *
+         * @param rvec 3x1 rotation vector. The direction indicates the rotation axis and
+         *             its length indicates the rotation angle in radian (using the right thumb convention).
+         */
+        void rotation(const Vec3& rvec);
+
+        /**
+         * Combines rotation methods above. Supports 3x3, 1x3, 3x1 sizes of data matrix.
+         *
+         * It sets the upper left 3x3 part of the matrix. The remaining part is unaffected.
+         *
+         * @param data 1-channel matrix.
+         *             When it is a 3x3 matrix, it sets the upper left 3x3 part of the current matrix.
+         *             When it is a 1x3 or 3x1 matrix, it is used as a rotation vector. The Rodrigues formula
+         *             is used to compute the rotation matrix and sets the upper left 3x3 part of the current matrix.
+         */
+        void rotation(const Mat& data);
+
+        /**
+         * Copy the 3x3 matrix L to the upper left part of the current matrix
+         *
+         * It sets the upper left 3x3 part of the matrix. The remaining part is unaffected.
+         *
+         * @param L 3x3 matrix.
+         */
+        void linear(const Mat3& L);
+
+        /**
+         * Copy t to the first three elements of the last column of the current matrix
+         *
+         * It sets the upper right 3x1 part of the matrix. The remaining part is unaffected.
+         *
+         * @param t 3x1 translation vector.
+         */
+        void translation(const Vec3& t);
+
+        //! @return the upper left 3x3 part
+        Mat3 rotation() const;
+
+        //! @return the upper left 3x3 part
+        Mat3 linear() const;
+
+        //! @return the upper right 3x1 part
+        Vec3 translation() const;
+
+        //! Rodrigues vector.
+        //! @return a vector representing the upper left 3x3 rotation matrix of the current matrix.
+        //! @warning  Since the mapping between rotation vectors and rotation matrices is many to one,
+        //!           this function returns only one rotation vector that represents the current rotation matrix,
+        //!           which is not necessarily the same one set by `rotation(const Vec3& rvec)`.
+        Vec3 rvec() const;
+
+        //! @return the inverse of the current matrix.
+        Affine3 inv(int method = cv::DECOMP_SVD) const;
+
+        //! a.rotate(R) is equivalent to Affine(R, 0) * a;
+        Affine3 rotate(const Mat3& R) const;
+
+        //! a.rotate(rvec) is equivalent to Affine(rvec, 0) * a;
+        Affine3 rotate(const Vec3& rvec) const;
+
+        //! a.translate(t) is equivalent to Affine(E, t) * a, where E is an identity matrix
+        Affine3 translate(const Vec3& t) const;
+
+        //! a.concatenate(affine) is equivalent to affine * a;
+        Affine3 concatenate(const Affine3& affine) const;
+
+        template <typename Y> operator Affine3<Y>() const;
+
+        template <typename Y> Affine3<Y> cast() const;
+
+        Mat4 matrix;
+
+#if defined EIGEN_WORLD_VERSION && defined EIGEN_GEOMETRY_MODULE_H
+        Affine3(const Eigen::Transform<T, 3, Eigen::Affine, (Eigen::RowMajor)>& affine);
+        Affine3(const Eigen::Transform<T, 3, Eigen::Affine>& affine);
+        operator Eigen::Transform<T, 3, Eigen::Affine, (Eigen::RowMajor)>() const;
+        operator Eigen::Transform<T, 3, Eigen::Affine>() const;
+#endif
+    };
+
+    template<typename T> static
+    Affine3<T> operator*(const Affine3<T>& affine1, const Affine3<T>& affine2);
+
+    //! V is a 3-element vector with member fields x, y and z
+    template<typename T, typename V> static
+    V operator*(const Affine3<T>& affine, const V& vector);
+
+    typedef Affine3<float> Affine3f;
+    typedef Affine3<double> Affine3d;
+
+    static Vec3f operator*(const Affine3f& affine, const Vec3f& vector);
+    static Vec3d operator*(const Affine3d& affine, const Vec3d& vector);
+
+    template<typename _Tp> class DataType< Affine3<_Tp> >
+    {
+    public:
+        typedef Affine3<_Tp>                               value_type;
+        typedef Affine3<typename DataType<_Tp>::work_type> work_type;
+        typedef _Tp                                        channel_type;
+
+        enum { generic_type = 0,
+               channels     = 16,
+               fmt          = traits::SafeFmt<channel_type>::fmt + ((channels - 1) << 8)
+#ifdef OPENCV_TRAITS_ENABLE_DEPRECATED
+               ,depth        = DataType<channel_type>::depth
+               ,type         = CV_MAKETYPE(depth, channels)
+#endif
+             };
+
+        typedef Vec<channel_type, channels> vec_type;
+    };
+
+    namespace traits {
+    template<typename _Tp>
+    struct Depth< Affine3<_Tp> > { enum { value = Depth<_Tp>::value }; };
+    template<typename _Tp>
+    struct Type< Affine3<_Tp> > { enum { value = CV_MAKETYPE(Depth<_Tp>::value, 16) }; };
+    } // namespace
+
+//! @} core
+
+}
+
+//! @cond IGNORED
+
+///////////////////////////////////////////////////////////////////////////////////
+// Implementation
+
+template<typename T> inline
+cv::Affine3<T>::Affine3()
+    : matrix(Mat4::eye())
+{}
+
+template<typename T> inline
+cv::Affine3<T>::Affine3(const Mat4& affine)
+    : matrix(affine)
+{}
+
+template<typename T> inline
+cv::Affine3<T>::Affine3(const Mat3& R, const Vec3& t)
+{
+    rotation(R);
+    translation(t);
+    matrix.val[12] = matrix.val[13] = matrix.val[14] = 0;
+    matrix.val[15] = 1;
+}
+
+template<typename T> inline
+cv::Affine3<T>::Affine3(const Vec3& _rvec, const Vec3& t)
+{
+    rotation(_rvec);
+    translation(t);
+    matrix.val[12] = matrix.val[13] = matrix.val[14] = 0;
+    matrix.val[15] = 1;
+}
+
+template<typename T> inline
+cv::Affine3<T>::Affine3(const cv::Mat& data, const Vec3& t)
+{
+    CV_Assert(data.type() == cv::traits::Type<T>::value);
+    CV_Assert(data.channels() == 1);
+
+    if (data.cols == 4 && data.rows == 4)
+    {
+        data.copyTo(matrix);
+        return;
+    }
+    else if (data.cols == 4 && data.rows == 3)
+    {
+        rotation(data(Rect(0, 0, 3, 3)));
+        translation(data(Rect(3, 0, 1, 3)));
+    }
+    else
+    {
+        rotation(data);
+        translation(t);
+    }
+
+    matrix.val[12] = matrix.val[13] = matrix.val[14] = 0;
+    matrix.val[15] = 1;
+}
+
+template<typename T> inline
+cv::Affine3<T>::Affine3(const float_type* vals) : matrix(vals)
+{}
+
+template<typename T> inline
+cv::Affine3<T> cv::Affine3<T>::Identity()
+{
+    return Affine3<T>(cv::Affine3<T>::Mat4::eye());
+}
+
+template<typename T> inline
+void cv::Affine3<T>::rotation(const Mat3& R)
+{
+    linear(R);
+}
+
+template<typename T> inline
+void cv::Affine3<T>::rotation(const Vec3& _rvec)
+{
+    double theta = norm(_rvec);
+
+    if (theta < DBL_EPSILON)
+        rotation(Mat3::eye());
+    else
+    {
+        double c = std::cos(theta);
+        double s = std::sin(theta);
+        double c1 = 1. - c;
+        double itheta = (theta != 0) ? 1./theta : 0.;
+
+        Point3_<T> r = _rvec*itheta;
+
+        Mat3 rrt( r.x*r.x, r.x*r.y, r.x*r.z, r.x*r.y, r.y*r.y, r.y*r.z, r.x*r.z, r.y*r.z, r.z*r.z );
+        Mat3 r_x( 0, -r.z, r.y, r.z, 0, -r.x, -r.y, r.x, 0 );
+
+        // R = cos(theta)*I + (1 - cos(theta))*r*rT + sin(theta)*[r_x]
+        // where [r_x] is [0 -rz ry; rz 0 -rx; -ry rx 0]
+        Mat3 R = c*Mat3::eye() + c1*rrt + s*r_x;
+
+        rotation(R);
+    }
+}
+
+//Combines rotation methods above. Supports 3x3, 1x3, 3x1 sizes of data matrix;
+template<typename T> inline
+void cv::Affine3<T>::rotation(const cv::Mat& data)
+{
+    CV_Assert(data.type() == cv::traits::Type<T>::value);
+    CV_Assert(data.channels() == 1);
+
+    if (data.cols == 3 && data.rows == 3)
+    {
+        Mat3 R;
+        data.copyTo(R);
+        rotation(R);
+    }
+    else if ((data.cols == 3 && data.rows == 1) || (data.cols == 1 && data.rows == 3))
+    {
+        Vec3 _rvec;
+        data.reshape(1, 3).copyTo(_rvec);
+        rotation(_rvec);
+    }
+    else
+        CV_Error(Error::StsError, "Input matrix can only be 3x3, 1x3 or 3x1");
+}
+
+template<typename T> inline
+void cv::Affine3<T>::linear(const Mat3& L)
+{
+    matrix.val[0] = L.val[0]; matrix.val[1] = L.val[1];  matrix.val[ 2] = L.val[2];
+    matrix.val[4] = L.val[3]; matrix.val[5] = L.val[4];  matrix.val[ 6] = L.val[5];
+    matrix.val[8] = L.val[6]; matrix.val[9] = L.val[7];  matrix.val[10] = L.val[8];
+}
+
+template<typename T> inline
+void cv::Affine3<T>::translation(const Vec3& t)
+{
+    matrix.val[3] = t[0]; matrix.val[7] = t[1]; matrix.val[11] = t[2];
+}
+
+template<typename T> inline
+typename cv::Affine3<T>::Mat3 cv::Affine3<T>::rotation() const
+{
+    return linear();
+}
+
+template<typename T> inline
+typename cv::Affine3<T>::Mat3 cv::Affine3<T>::linear() const
+{
+    typename cv::Affine3<T>::Mat3 R;
+    R.val[0] = matrix.val[0];  R.val[1] = matrix.val[1];  R.val[2] = matrix.val[ 2];
+    R.val[3] = matrix.val[4];  R.val[4] = matrix.val[5];  R.val[5] = matrix.val[ 6];
+    R.val[6] = matrix.val[8];  R.val[7] = matrix.val[9];  R.val[8] = matrix.val[10];
+    return R;
+}
+
+template<typename T> inline
+typename cv::Affine3<T>::Vec3 cv::Affine3<T>::translation() const
+{
+    return Vec3(matrix.val[3], matrix.val[7], matrix.val[11]);
+}
+
+template<typename T> inline
+typename cv::Affine3<T>::Vec3 cv::Affine3<T>::rvec() const
+{
+    cv::Vec3d w;
+    cv::Matx33d u, vt, R = rotation();
+    cv::SVD::compute(R, w, u, vt, cv::SVD::FULL_UV + cv::SVD::MODIFY_A);
+    R = u * vt;
+
+    double rx = R.val[7] - R.val[5];
+    double ry = R.val[2] - R.val[6];
+    double rz = R.val[3] - R.val[1];
+
+    double s = std::sqrt((rx*rx + ry*ry + rz*rz)*0.25);
+    double c = (R.val[0] + R.val[4] + R.val[8] - 1) * 0.5;
+    c = c > 1.0 ? 1.0 : c < -1.0 ? -1.0 : c;
+    double theta = acos(c);
+
+    if( s < 1e-5 )
+    {
+        if( c > 0 )
+            rx = ry = rz = 0;
+        else
+        {
+            double t;
+            t = (R.val[0] + 1) * 0.5;
+            rx = std::sqrt(std::max(t, 0.0));
+            t = (R.val[4] + 1) * 0.5;
+            ry = std::sqrt(std::max(t, 0.0)) * (R.val[1] < 0 ? -1.0 : 1.0);
+            t = (R.val[8] + 1) * 0.5;
+            rz = std::sqrt(std::max(t, 0.0)) * (R.val[2] < 0 ? -1.0 : 1.0);
+
+            if( fabs(rx) < fabs(ry) && fabs(rx) < fabs(rz) && (R.val[5] > 0) != (ry*rz > 0) )
+                rz = -rz;
+            theta /= std::sqrt(rx*rx + ry*ry + rz*rz);
+            rx *= theta;
+            ry *= theta;
+            rz *= theta;
+        }
+    }
+    else
+    {
+        double vth = 1/(2*s);
+        vth *= theta;
+        rx *= vth; ry *= vth; rz *= vth;
+    }
+
+    return cv::Vec3d(rx, ry, rz);
+}
+
+template<typename T> inline
+cv::Affine3<T> cv::Affine3<T>::inv(int method) const
+{
+    return matrix.inv(method);
+}
+
+template<typename T> inline
+cv::Affine3<T> cv::Affine3<T>::rotate(const Mat3& R) const
+{
+    Mat3 Lc = linear();
+    Vec3 tc = translation();
+    Mat4 result;
+    result.val[12] = result.val[13] = result.val[14] = 0;
+    result.val[15] = 1;
+
+    for(int j = 0; j < 3; ++j)
+    {
+        for(int i = 0; i < 3; ++i)
+        {
+            float_type value = 0;
+            for(int k = 0; k < 3; ++k)
+                value += R(j, k) * Lc(k, i);
+            result(j, i) = value;
+        }
+
+        result(j, 3) = R.row(j).dot(tc.t());
+    }
+    return result;
+}
+
+template<typename T> inline
+cv::Affine3<T> cv::Affine3<T>::rotate(const Vec3& _rvec) const
+{
+    return rotate(Affine3f(_rvec).rotation());
+}
+
+template<typename T> inline
+cv::Affine3<T> cv::Affine3<T>::translate(const Vec3& t) const
+{
+    Mat4 m = matrix;
+    m.val[ 3] += t[0];
+    m.val[ 7] += t[1];
+    m.val[11] += t[2];
+    return m;
+}
+
+template<typename T> inline
+cv::Affine3<T> cv::Affine3<T>::concatenate(const Affine3<T>& affine) const
+{
+    return (*this).rotate(affine.rotation()).translate(affine.translation());
+}
+
+template<typename T> template <typename Y> inline
+cv::Affine3<T>::operator Affine3<Y>() const
+{
+    return Affine3<Y>(matrix);
+}
+
+template<typename T> template <typename Y> inline
+cv::Affine3<Y> cv::Affine3<T>::cast() const
+{
+    return Affine3<Y>(matrix);
+}
+
+template<typename T> inline
+cv::Affine3<T> cv::operator*(const cv::Affine3<T>& affine1, const cv::Affine3<T>& affine2)
+{
+    return affine2.concatenate(affine1);
+}
+
+template<typename T, typename V> inline
+V cv::operator*(const cv::Affine3<T>& affine, const V& v)
+{
+    const typename Affine3<T>::Mat4& m = affine.matrix;
+
+    V r;
+    r.x = m.val[0] * v.x + m.val[1] * v.y + m.val[ 2] * v.z + m.val[ 3];
+    r.y = m.val[4] * v.x + m.val[5] * v.y + m.val[ 6] * v.z + m.val[ 7];
+    r.z = m.val[8] * v.x + m.val[9] * v.y + m.val[10] * v.z + m.val[11];
+    return r;
+}
+
+static inline
+cv::Vec3f cv::operator*(const cv::Affine3f& affine, const cv::Vec3f& v)
+{
+    const cv::Matx44f& m = affine.matrix;
+    cv::Vec3f r;
+    r.val[0] = m.val[0] * v[0] + m.val[1] * v[1] + m.val[ 2] * v[2] + m.val[ 3];
+    r.val[1] = m.val[4] * v[0] + m.val[5] * v[1] + m.val[ 6] * v[2] + m.val[ 7];
+    r.val[2] = m.val[8] * v[0] + m.val[9] * v[1] + m.val[10] * v[2] + m.val[11];
+    return r;
+}
+
+static inline
+cv::Vec3d cv::operator*(const cv::Affine3d& affine, const cv::Vec3d& v)
+{
+    const cv::Matx44d& m = affine.matrix;
+    cv::Vec3d r;
+    r.val[0] = m.val[0] * v[0] + m.val[1] * v[1] + m.val[ 2] * v[2] + m.val[ 3];
+    r.val[1] = m.val[4] * v[0] + m.val[5] * v[1] + m.val[ 6] * v[2] + m.val[ 7];
+    r.val[2] = m.val[8] * v[0] + m.val[9] * v[1] + m.val[10] * v[2] + m.val[11];
+    return r;
+}
+
+
+
+#if defined EIGEN_WORLD_VERSION && defined EIGEN_GEOMETRY_MODULE_H
+
+template<typename T> inline
+cv::Affine3<T>::Affine3(const Eigen::Transform<T, 3, Eigen::Affine, (Eigen::RowMajor)>& affine)
+{
+    cv::Mat(4, 4, cv::traits::Type<T>::value, affine.matrix().data()).copyTo(matrix);
+}
+
+template<typename T> inline
+cv::Affine3<T>::Affine3(const Eigen::Transform<T, 3, Eigen::Affine>& affine)
+{
+    Eigen::Transform<T, 3, Eigen::Affine, (Eigen::RowMajor)> a = affine;
+    cv::Mat(4, 4, cv::traits::Type<T>::value, a.matrix().data()).copyTo(matrix);
+}
+
+template<typename T> inline
+cv::Affine3<T>::operator Eigen::Transform<T, 3, Eigen::Affine, (Eigen::RowMajor)>() const
+{
+    Eigen::Transform<T, 3, Eigen::Affine, (Eigen::RowMajor)> r;
+    cv::Mat hdr(4, 4, cv::traits::Type<T>::value, r.matrix().data());
+    cv::Mat(matrix, false).copyTo(hdr);
+    return r;
+}
+
+template<typename T> inline
+cv::Affine3<T>::operator Eigen::Transform<T, 3, Eigen::Affine>() const
+{
+    return this->operator Eigen::Transform<T, 3, Eigen::Affine, (Eigen::RowMajor)>();
+}
+
+#endif /* defined EIGEN_WORLD_VERSION && defined EIGEN_GEOMETRY_MODULE_H */
+
+//! @endcond
+
+#endif /* __cplusplus */
+
+#endif /* OPENCV_CORE_AFFINE3_HPP */

model_zoo/official/lite/app/src/main/cpp/include/opencv2/core/async.hpp

@@ -0,0 +1,105 @@
+// This file is part of OpenCV project.
+// It is subject to the license terms in the LICENSE file found in the top-level directory
+// of this distribution and at http://opencv.org/license.html.
+
+#ifndef OPENCV_CORE_ASYNC_HPP
+#define OPENCV_CORE_ASYNC_HPP
+
+#include <opencv2/core/mat.hpp>
+
+#ifdef CV_CXX11
+//#include <future>
+#include <chrono>
+#endif
+
+namespace cv {
+
+/** @addtogroup core_async
+
+@{
+*/
+
+
+/** @brief Returns result of asynchronous operations
+
+Object has attached asynchronous state.
+Assignment operator doesn't clone asynchronous state (it is shared between all instances).
+
+Result can be fetched via get() method only once.
+
+*/
+class CV_EXPORTS_W AsyncArray
+{
+public:
+    ~AsyncArray() CV_NOEXCEPT;
+    CV_WRAP AsyncArray() CV_NOEXCEPT;
+    AsyncArray(const AsyncArray& o) CV_NOEXCEPT;
+    AsyncArray& operator=(const AsyncArray& o) CV_NOEXCEPT;
+    CV_WRAP void release() CV_NOEXCEPT;
+
+    /** Fetch the result.
+    @param[out] dst destination array
+
+    Waits for result until container has valid result.
+    Throws exception if exception was stored as a result.
+
+    Throws exception on invalid container state.
+
+    @note Result or stored exception can be fetched only once.
+    */
+    CV_WRAP void get(OutputArray dst) const;
+
+    /** Retrieving the result with timeout
+    @param[out] dst destination array
+    @param[in] timeoutNs timeout in nanoseconds, -1 for infinite wait
+
+    @returns true if result is ready, false if the timeout has expired
+
+    @note Result or stored exception can be fetched only once.
+    */
+    bool get(OutputArray dst, int64 timeoutNs) const;
+
+    CV_WRAP inline
+    bool get(OutputArray dst, double timeoutNs) const { return get(dst, (int64)timeoutNs); }
+
+    bool wait_for(int64 timeoutNs) const;
+
+    CV_WRAP inline
+    bool wait_for(double timeoutNs) const { return wait_for((int64)timeoutNs); }
+
+    CV_WRAP bool valid() const CV_NOEXCEPT;
+
+#ifdef CV_CXX11
+    inline AsyncArray(AsyncArray&& o) { p = o.p; o.p = NULL; }
+    inline AsyncArray& operator=(AsyncArray&& o) CV_NOEXCEPT { std::swap(p, o.p); return *this; }
+
+    template<typename _Rep, typename _Period>
+    inline bool get(OutputArray dst, const std::chrono::duration<_Rep, _Period>& timeout)
+    {
+        return get(dst, (int64)(std::chrono::nanoseconds(timeout).count()));
+    }
+
+    template<typename _Rep, typename _Period>
+    inline bool wait_for(const std::chrono::duration<_Rep, _Period>& timeout)
+    {
+        return wait_for((int64)(std::chrono::nanoseconds(timeout).count()));
+    }
+
+#if 0
+    std::future<Mat> getFutureMat() const;
+    std::future<UMat> getFutureUMat() const;
+#endif
+#endif
+
+
+    // PImpl
+    struct Impl; friend struct Impl;
+    inline void* _getImpl() const CV_NOEXCEPT { return p; }
+protected:
+    Impl* p;
+};
+
+
+//! @}
+} // namespace
+#endif // OPENCV_CORE_ASYNC_HPP

model_zoo/official/lite/app/src/main/cpp/include/opencv2/core/base.hpp

@@ -0,0 +1,654 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Copyright (C) 2014, Itseez Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CORE_BASE_HPP
+#define OPENCV_CORE_BASE_HPP
+
+#ifndef __cplusplus
+#  error base.hpp header must be compiled as C++
+#endif
+
+#include "opencv2/opencv_modules.hpp"
+
+#include <climits>
+#include <algorithm>
+
+#include "opencv2/core/cvdef.h"
+#include "opencv2/core/cvstd.hpp"
+
+namespace cv
+{
+
+//! @addtogroup core_utils
+//! @{
+
+namespace Error {
+//! error codes
+enum Code {
+    StsOk=                       0,  //!< everything is ok
+    StsBackTrace=               -1,  //!< pseudo error for back trace
+    StsError=                   -2,  //!< unknown /unspecified error
+    StsInternal=                -3,  //!< internal error (bad state)
+    StsNoMem=                   -4,  //!< insufficient memory
+    StsBadArg=                  -5,  //!< function arg/param is bad
+    StsBadFunc=                 -6,  //!< unsupported function
+    StsNoConv=                  -7,  //!< iteration didn't converge
+    StsAutoTrace=               -8,  //!< tracing
+    HeaderIsNull=               -9,  //!< image header is NULL
+    BadImageSize=              -10,  //!< image size is invalid
+    BadOffset=                 -11,  //!< offset is invalid
+    BadDataPtr=                -12,  //!<
+    BadStep=                   -13,  //!< image step is wrong, this may happen for a non-continuous matrix.
+    BadModelOrChSeq=           -14,  //!<
+    BadNumChannels=            -15,  //!< bad number of channels, for example, some functions accept only single channel matrices.
+    BadNumChannel1U=           -16,  //!<
+    BadDepth=                  -17,  //!< input image depth is not supported by the function
+    BadAlphaChannel=           -18,  //!<
+    BadOrder=                  -19,  //!< number of dimensions is out of range
+    BadOrigin=                 -20,  //!< incorrect input origin
+    BadAlign=                  -21,  //!< incorrect input align
+    BadCallBack=               -22,  //!<
+    BadTileSize=               -23,  //!<
+    BadCOI=                    -24,  //!< input COI is not supported
+    BadROISize=                -25,  //!< incorrect input roi
+    MaskIsTiled=               -26,  //!<
+    StsNullPtr=                -27,  //!< null pointer
+    StsVecLengthErr=           -28,  //!< incorrect vector length
+    StsFilterStructContentErr= -29,  //!< incorrect filter structure content
+    StsKernelStructContentErr= -30,  //!< incorrect transform kernel content
+    StsFilterOffsetErr=        -31,  //!< incorrect filter offset value
+    StsBadSize=                -201, //!< the input/output structure size is incorrect
+    StsDivByZero=              -202, //!< division by zero
+    StsInplaceNotSupported=    -203, //!< in-place operation is not supported
+    StsObjectNotFound=         -204, //!< request can't be completed
+    StsUnmatchedFormats=       -205, //!< formats of input/output arrays differ
+    StsBadFlag=                -206, //!< flag is wrong or not supported
+    StsBadPoint=               -207, //!< bad CvPoint
+    StsBadMask=                -208, //!< bad format of mask (neither 8uC1 nor 8sC1)
+    StsUnmatchedSizes=         -209, //!< sizes of input/output structures do not match
+    StsUnsupportedFormat=      -210, //!< the data format/type is not supported by the function
+    StsOutOfRange=             -211, //!< some of parameters are out of range
+    StsParseError=             -212, //!< invalid syntax/structure of the parsed file
+    StsNotImplemented=         -213, //!< the requested function/feature is not implemented
+    StsBadMemBlock=            -214, //!< an allocated block has been corrupted
+    StsAssert=                 -215, //!< assertion failed
+    GpuNotSupported=           -216, //!< no CUDA support
+    GpuApiCallError=           -217, //!< GPU API call error
+    OpenGlNotSupported=        -218, //!< no OpenGL support
+    OpenGlApiCallError=        -219, //!< OpenGL API call error
+    OpenCLApiCallError=        -220, //!< OpenCL API call error
+    OpenCLDoubleNotSupported=  -221,
+    OpenCLInitError=           -222, //!< OpenCL initialization error
+    OpenCLNoAMDBlasFft=        -223
+};
+} //Error
+
+//! @} core_utils
+
+//! @addtogroup core_array
+//! @{
+
+//! matrix decomposition types
+enum DecompTypes {
+    /** Gaussian elimination with the optimal pivot element chosen. */
+    DECOMP_LU       = 0,
+    /** singular value decomposition (SVD) method; the system can be over-defined and/or the matrix
+    src1 can be singular */
+    DECOMP_SVD      = 1,
+    /** eigenvalue decomposition; the matrix src1 must be symmetrical */
+    DECOMP_EIG      = 2,
+    /** Cholesky \f$LL^T\f$ factorization; the matrix src1 must be symmetrical and positively
+    defined */
+    DECOMP_CHOLESKY = 3,
+    /** QR factorization; the system can be over-defined and/or the matrix src1 can be singular */
+    DECOMP_QR       = 4,
+    /** while all the previous flags are mutually exclusive, this flag can be used together with
+    any of the previous; it means that the normal equations
+    \f$\texttt{src1}^T\cdot\texttt{src1}\cdot\texttt{dst}=\texttt{src1}^T\texttt{src2}\f$ are
+    solved instead of the original system
+    \f$\texttt{src1}\cdot\texttt{dst}=\texttt{src2}\f$ */
+    DECOMP_NORMAL   = 16
+};
+
+/** norm types
+
+src1 and src2 denote input arrays.
+*/
+
+enum NormTypes {
+                /**
+                \f[
+                norm =  \forkthree
+                {\|\texttt{src1}\|_{L_{\infty}} =  \max _I | \texttt{src1} (I)|}{if  \(\texttt{normType} = \texttt{NORM_INF}\) }
+                {\|\texttt{src1}-\texttt{src2}\|_{L_{\infty}} =  \max _I | \texttt{src1} (I) -  \texttt{src2} (I)|}{if  \(\texttt{normType} = \texttt{NORM_INF}\) }
+                {\frac{\|\texttt{src1}-\texttt{src2}\|_{L_{\infty}}    }{\|\texttt{src2}\|_{L_{\infty}} }}{if  \(\texttt{normType} = \texttt{NORM_RELATIVE | NORM_INF}\) }
+                \f]
+                */
+                NORM_INF       = 1,
+                /**
+                \f[
+                norm =  \forkthree
+                {\| \texttt{src1} \| _{L_1} =  \sum _I | \texttt{src1} (I)|}{if  \(\texttt{normType} = \texttt{NORM_L1}\)}
+                { \| \texttt{src1} - \texttt{src2} \| _{L_1} =  \sum _I | \texttt{src1} (I) -  \texttt{src2} (I)|}{if  \(\texttt{normType} = \texttt{NORM_L1}\) }
+                { \frac{\|\texttt{src1}-\texttt{src2}\|_{L_1} }{\|\texttt{src2}\|_{L_1}} }{if  \(\texttt{normType} = \texttt{NORM_RELATIVE | NORM_L1}\) }
+                \f]*/
+                 NORM_L1        = 2,
+                 /**
+                 \f[
+                 norm =  \forkthree
+                 { \| \texttt{src1} \| _{L_2} =  \sqrt{\sum_I \texttt{src1}(I)^2} }{if  \(\texttt{normType} = \texttt{NORM_L2}\) }
+                 { \| \texttt{src1} - \texttt{src2} \| _{L_2} =  \sqrt{\sum_I (\texttt{src1}(I) - \texttt{src2}(I))^2} }{if  \(\texttt{normType} = \texttt{NORM_L2}\) }
+                 { \frac{\|\texttt{src1}-\texttt{src2}\|_{L_2} }{\|\texttt{src2}\|_{L_2}} }{if  \(\texttt{normType} = \texttt{NORM_RELATIVE | NORM_L2}\) }
+                 \f]
+                 */
+                 NORM_L2        = 4,
+                 /**
+                 \f[
+                 norm =  \forkthree
+                 { \| \texttt{src1} \| _{L_2} ^{2} = \sum_I \texttt{src1}(I)^2} {if  \(\texttt{normType} = \texttt{NORM_L2SQR}\)}
+                 { \| \texttt{src1} - \texttt{src2} \| _{L_2} ^{2} =  \sum_I (\texttt{src1}(I) - \texttt{src2}(I))^2 }{if  \(\texttt{normType} = \texttt{NORM_L2SQR}\) }
+                 { \left(\frac{\|\texttt{src1}-\texttt{src2}\|_{L_2} }{\|\texttt{src2}\|_{L_2}}\right)^2 }{if  \(\texttt{normType} = \texttt{NORM_RELATIVE | NORM_L2SQR}\) }
+                 \f]
+                 */
+                 NORM_L2SQR     = 5,
+                 /**
+                 In the case of one input array, calculates the Hamming distance of the array from zero,
+                 In the case of two input arrays, calculates the Hamming distance between the arrays.
+                 */
+                 NORM_HAMMING   = 6,
+                 /**
+                 Similar to NORM_HAMMING, but in the calculation, each two bits of the input sequence will
+                 be added and treated as a single bit to be used in the same calculation as NORM_HAMMING.
+                 */
+                 NORM_HAMMING2  = 7,
+                 NORM_TYPE_MASK = 7, //!< bit-mask which can be used to separate norm type from norm flags
+                 NORM_RELATIVE  = 8, //!< flag
+                 NORM_MINMAX    = 32 //!< flag
+               };
+
+//! comparison types
+enum CmpTypes { CMP_EQ = 0, //!< src1 is equal to src2.
+                CMP_GT = 1, //!< src1 is greater than src2.
+                CMP_GE = 2, //!< src1 is greater than or equal to src2.
+                CMP_LT = 3, //!< src1 is less than src2.
+                CMP_LE = 4, //!< src1 is less than or equal to src2.
+                CMP_NE = 5  //!< src1 is unequal to src2.
+              };
+
+//! generalized matrix multiplication flags
+enum GemmFlags { GEMM_1_T = 1, //!< transposes src1
+                 GEMM_2_T = 2, //!< transposes src2
+                 GEMM_3_T = 4 //!< transposes src3
+               };
+
+enum DftFlags {
+    /** performs an inverse 1D or 2D transform instead of the default forward
+        transform. */
+    DFT_INVERSE        = 1,
+    /** scales the result: divide it by the number of array elements. Normally, it is
+        combined with DFT_INVERSE. */
+    DFT_SCALE          = 2,
+    /** performs a forward or inverse transform of every individual row of the input
+        matrix; this flag enables you to transform multiple vectors simultaneously and can be used to
+        decrease the overhead (which is sometimes several times larger than the processing itself) to
+        perform 3D and higher-dimensional transformations and so forth.*/
+    DFT_ROWS           = 4,
+    /** performs a forward transformation of 1D or 2D real array; the result,
+        though being a complex array, has complex-conjugate symmetry (*CCS*, see the function
+        description below for details), and such an array can be packed into a real array of the same
+        size as input, which is the fastest option and which is what the function does by default;
+        however, you may wish to get a full complex array (for simpler spectrum analysis, and so on) -
+        pass the flag to enable the function to produce a full-size complex output array. */
+    DFT_COMPLEX_OUTPUT = 16,
+    /** performs an inverse transformation of a 1D or 2D complex array; the
+        result is normally a complex array of the same size, however, if the input array has
+        conjugate-complex symmetry (for example, it is a result of forward transformation with
+        DFT_COMPLEX_OUTPUT flag), the output is a real array; while the function itself does not
+        check whether the input is symmetrical or not, you can pass the flag and then the function
+        will assume the symmetry and produce the real output array (note that when the input is packed
+        into a real array and inverse transformation is executed, the function treats the input as a
+        packed complex-conjugate symmetrical array, and the output will also be a real array). */
+    DFT_REAL_OUTPUT    = 32,
+    /** specifies that input is complex input. If this flag is set, the input must have 2 channels.
+        On the other hand, for backwards compatibility reason, if input has 2 channels, input is
+        already considered complex. */
+    DFT_COMPLEX_INPUT  = 64,
+    /** performs an inverse 1D or 2D transform instead of the default forward transform. */
+    DCT_INVERSE        = DFT_INVERSE,
+    /** performs a forward or inverse transform of every individual row of the input
+        matrix. This flag enables you to transform multiple vectors simultaneously and can be used to
+        decrease the overhead (which is sometimes several times larger than the processing itself) to
+        perform 3D and higher-dimensional transforms and so forth.*/
+    DCT_ROWS           = DFT_ROWS
+};
+
+//! Various border types, image boundaries are denoted with `|`
+//! @see borderInterpolate, copyMakeBorder
+enum BorderTypes {
+    BORDER_CONSTANT    = 0, //!< `iiiiii|abcdefgh|iiiiiii`  with some specified `i`
+    BORDER_REPLICATE   = 1, //!< `aaaaaa|abcdefgh|hhhhhhh`
+    BORDER_REFLECT     = 2, //!< `fedcba|abcdefgh|hgfedcb`
+    BORDER_WRAP        = 3, //!< `cdefgh|abcdefgh|abcdefg`
+    BORDER_REFLECT_101 = 4, //!< `gfedcb|abcdefgh|gfedcba`
+    BORDER_TRANSPARENT = 5, //!< `uvwxyz|abcdefgh|ijklmno`
+
+    BORDER_REFLECT101  = BORDER_REFLECT_101, //!< same as BORDER_REFLECT_101
+    BORDER_DEFAULT     = BORDER_REFLECT_101, //!< same as BORDER_REFLECT_101
+    BORDER_ISOLATED    = 16 //!< do not look outside of ROI
+};
+
+//! @} core_array
+
+//! @addtogroup core_utils
+//! @{
+
+/*! @brief Signals an error and raises the exception.
+
+By default the function prints information about the error to stderr,
+then it either stops if setBreakOnError() had been called before or raises the exception.
+It is possible to alternate error processing by using redirectError().
+@param _code - error code (Error::Code)
+@param _err - error description
+@param _func - function name. Available only when the compiler supports getting it
+@param _file - source file name where the error has occurred
+@param _line - line number in the source file where the error has occurred
+@see CV_Error, CV_Error_, CV_Assert, CV_DbgAssert
+ */
+CV_EXPORTS CV_NORETURN void error(int _code, const String& _err, const char* _func, const char* _file, int _line);
+
+#ifdef CV_STATIC_ANALYSIS
+
+// In practice, some macro are not processed correctly (noreturn is not detected).
+// We need to use simplified definition for them.
+#define CV_Error(code, msg) do { (void)(code); (void)(msg); abort(); } while (0)
+#define CV_Error_(code, args) do { (void)(code); (void)(cv::format args); abort(); } while (0)
+#define CV_Assert( expr ) do { if (!(expr)) abort(); } while (0)
+
+#else // CV_STATIC_ANALYSIS
+
+/** @brief Call the error handler.
+
+Currently, the error handler prints the error code and the error message to the standard
+error stream `stderr`. In the Debug configuration, it then provokes memory access violation, so that
+the execution stack and all the parameters can be analyzed by the debugger. In the Release
+configuration, the exception is thrown.
+
+@param code one of Error::Code
+@param msg error message
+*/
+#define CV_Error( code, msg ) cv::error( code, msg, CV_Func, __FILE__, __LINE__ )
+
+/**  @brief Call the error handler.
+
+This macro can be used to construct an error message on-fly to include some dynamic information,
+for example:
+@code
+    // note the extra parentheses around the formatted text message
+    CV_Error_(Error::StsOutOfRange,
+    ("the value at (%d, %d)=%g is out of range", badPt.x, badPt.y, badValue));
+@endcode
+@param code one of Error::Code
+@param args printf-like formatted error message in parentheses
+*/
+#define CV_Error_( code, args ) cv::error( code, cv::format args, CV_Func, __FILE__, __LINE__ )
+
+/** @brief Checks a condition at runtime and throws exception if it fails
+
+The macros CV_Assert (and CV_DbgAssert(expr)) evaluate the specified expression. If it is 0, the macros
+raise an error (see cv::error). The macro CV_Assert checks the condition in both Debug and Release
+configurations while CV_DbgAssert is only retained in the Debug configuration.
+*/
+#define CV_Assert( expr ) do { if(!!(expr)) ; else cv::error( cv::Error::StsAssert, #expr, CV_Func, __FILE__, __LINE__ ); } while(0)
+
+#endif // CV_STATIC_ANALYSIS
+
+//! @cond IGNORED
+#if !defined(__OPENCV_BUILD)  // TODO: backward compatibility only
+#ifndef CV_ErrorNoReturn
+#define CV_ErrorNoReturn CV_Error
+#endif
+#ifndef CV_ErrorNoReturn_
+#define CV_ErrorNoReturn_ CV_Error_
+#endif
+#endif
+
+#define CV_Assert_1 CV_Assert
+#define CV_Assert_2( expr, ... ) CV_Assert_1(expr); __CV_EXPAND(CV_Assert_1( __VA_ARGS__ ))
+#define CV_Assert_3( expr, ... ) CV_Assert_1(expr); __CV_EXPAND(CV_Assert_2( __VA_ARGS__ ))
+#define CV_Assert_4( expr, ... ) CV_Assert_1(expr); __CV_EXPAND(CV_Assert_3( __VA_ARGS__ ))
+#define CV_Assert_5( expr, ... ) CV_Assert_1(expr); __CV_EXPAND(CV_Assert_4( __VA_ARGS__ ))
+#define CV_Assert_6( expr, ... ) CV_Assert_1(expr); __CV_EXPAND(CV_Assert_5( __VA_ARGS__ ))
+#define CV_Assert_7( expr, ... ) CV_Assert_1(expr); __CV_EXPAND(CV_Assert_6( __VA_ARGS__ ))
+#define CV_Assert_8( expr, ... ) CV_Assert_1(expr); __CV_EXPAND(CV_Assert_7( __VA_ARGS__ ))
+#define CV_Assert_9( expr, ... ) CV_Assert_1(expr); __CV_EXPAND(CV_Assert_8( __VA_ARGS__ ))
+#define CV_Assert_10( expr, ... ) CV_Assert_1(expr); __CV_EXPAND(CV_Assert_9( __VA_ARGS__ ))
+
+#define CV_Assert_N(...) do { __CV_EXPAND(__CV_CAT(CV_Assert_, __CV_VA_NUM_ARGS(__VA_ARGS__)) (__VA_ARGS__)); } while(0)
+
+//! @endcond
+
+#if defined _DEBUG || defined CV_STATIC_ANALYSIS
+#  define CV_DbgAssert(expr) CV_Assert(expr)
+#else
+/** replaced with CV_Assert(expr) in Debug configuration */
+#  define CV_DbgAssert(expr)
+#endif
+
+/*
+ * Hamming distance functor - counts the bit differences between two strings - useful for the Brief descriptor
+ * bit count of A exclusive XOR'ed with B
+ */
+struct CV_EXPORTS Hamming
+{
+    static const NormTypes normType = NORM_HAMMING;
+    typedef unsigned char ValueType;
+    typedef int ResultType;
+
+    /** this will count the bits in a ^ b
+     */
+    ResultType operator()( const unsigned char* a, const unsigned char* b, int size ) const;
+};
+
+typedef Hamming HammingLUT;
+
+/////////////////////////////////// inline norms ////////////////////////////////////
+
+template<typename _Tp> inline _Tp cv_abs(_Tp x) { return std::abs(x); }
+inline int cv_abs(uchar x) { return x; }
+inline int cv_abs(schar x) { return std::abs(x); }
+inline int cv_abs(ushort x) { return x; }
+inline int cv_abs(short x) { return std::abs(x); }
+
+template<typename _Tp, typename _AccTp> static inline
+_AccTp normL2Sqr(const _Tp* a, int n)
+{
+    _AccTp s = 0;
+    int i=0;
+#if CV_ENABLE_UNROLLED
+    for( ; i <= n - 4; i += 4 )
+    {
+        _AccTp v0 = a[i], v1 = a[i+1], v2 = a[i+2], v3 = a[i+3];
+        s += v0*v0 + v1*v1 + v2*v2 + v3*v3;
+    }
+#endif
+    for( ; i < n; i++ )
+    {
+        _AccTp v = a[i];
+        s += v*v;
+    }
+    return s;
+}
+
+template<typename _Tp, typename _AccTp> static inline
+_AccTp normL1(const _Tp* a, int n)
+{
+    _AccTp s = 0;
+    int i = 0;
+#if CV_ENABLE_UNROLLED
+    for(; i <= n - 4; i += 4 )
+    {
+        s += (_AccTp)cv_abs(a[i]) + (_AccTp)cv_abs(a[i+1]) +
+            (_AccTp)cv_abs(a[i+2]) + (_AccTp)cv_abs(a[i+3]);
+    }
+#endif
+    for( ; i < n; i++ )
+        s += cv_abs(a[i]);
+    return s;
+}
+
+template<typename _Tp, typename _AccTp> static inline
+_AccTp normInf(const _Tp* a, int n)
+{
+    _AccTp s = 0;
+    for( int i = 0; i < n; i++ )
+        s = std::max(s, (_AccTp)cv_abs(a[i]));
+    return s;
+}
+
+template<typename _Tp, typename _AccTp> static inline
+_AccTp normL2Sqr(const _Tp* a, const _Tp* b, int n)
+{
+    _AccTp s = 0;
+    int i= 0;
+#if CV_ENABLE_UNROLLED
+    for(; i <= n - 4; i += 4 )
+    {
+        _AccTp v0 = _AccTp(a[i] - b[i]), v1 = _AccTp(a[i+1] - b[i+1]), v2 = _AccTp(a[i+2] - b[i+2]), v3 = _AccTp(a[i+3] - b[i+3]);
+        s += v0*v0 + v1*v1 + v2*v2 + v3*v3;
+    }
+#endif
+    for( ; i < n; i++ )
+    {
+        _AccTp v = _AccTp(a[i] - b[i]);
+        s += v*v;
+    }
+    return s;
+}
+
+static inline float normL2Sqr(const float* a, const float* b, int n)
+{
+    float s = 0.f;
+    for( int i = 0; i < n; i++ )
+    {
+        float v = a[i] - b[i];
+        s += v*v;
+    }
+    return s;
+}
+
+template<typename _Tp, typename _AccTp> static inline
+_AccTp normL1(const _Tp* a, const _Tp* b, int n)
+{
+    _AccTp s = 0;
+    int i= 0;
+#if CV_ENABLE_UNROLLED
+    for(; i <= n - 4; i += 4 )
+    {
+        _AccTp v0 = _AccTp(a[i] - b[i]), v1 = _AccTp(a[i+1] - b[i+1]), v2 = _AccTp(a[i+2] - b[i+2]), v3 = _AccTp(a[i+3] - b[i+3]);
+        s += std::abs(v0) + std::abs(v1) + std::abs(v2) + std::abs(v3);
+    }
+#endif
+    for( ; i < n; i++ )
+    {
+        _AccTp v = _AccTp(a[i] - b[i]);
+        s += std::abs(v);
+    }
+    return s;
+}
+
+inline float normL1(const float* a, const float* b, int n)
+{
+    float s = 0.f;
+    for( int i = 0; i < n; i++ )
+    {
+        s += std::abs(a[i] - b[i]);
+    }
+    return s;
+}
+
+inline int normL1(const uchar* a, const uchar* b, int n)
+{
+    int s = 0;
+    for( int i = 0; i < n; i++ )
+    {
+        s += std::abs(a[i] - b[i]);
+    }
+    return s;
+}
+
+template<typename _Tp, typename _AccTp> static inline
+_AccTp normInf(const _Tp* a, const _Tp* b, int n)
+{
+    _AccTp s = 0;
+    for( int i = 0; i < n; i++ )
+    {
+        _AccTp v0 = a[i] - b[i];
+        s = std::max(s, std::abs(v0));
+    }
+    return s;
+}
+
+/** @brief Computes the cube root of an argument.
+
+ The function cubeRoot computes \f$\sqrt[3]{\texttt{val}}\f$. Negative arguments are handled correctly.
+ NaN and Inf are not handled. The accuracy approaches the maximum possible accuracy for
+ single-precision data.
+ @param val A function argument.
+ */
+CV_EXPORTS_W float cubeRoot(float val);
+
+/** @brief Calculates the angle of a 2D vector in degrees.
+
+ The function fastAtan2 calculates the full-range angle of an input 2D vector. The angle is measured
+ in degrees and varies from 0 to 360 degrees. The accuracy is about 0.3 degrees.
+ @param x x-coordinate of the vector.
+ @param y y-coordinate of the vector.
+ */
+CV_EXPORTS_W float fastAtan2(float y, float x);
+
+/** proxy for hal::LU */
+CV_EXPORTS int LU(float* A, size_t astep, int m, float* b, size_t bstep, int n);
+/** proxy for hal::LU */
+CV_EXPORTS int LU(double* A, size_t astep, int m, double* b, size_t bstep, int n);
+/** proxy for hal::Cholesky */
+CV_EXPORTS bool Cholesky(float* A, size_t astep, int m, float* b, size_t bstep, int n);
+/** proxy for hal::Cholesky */
+CV_EXPORTS bool Cholesky(double* A, size_t astep, int m, double* b, size_t bstep, int n);
+
+////////////////// forward declarations for important OpenCV types //////////////////
+
+//! @cond IGNORED
+
+template<typename _Tp, int cn> class Vec;
+template<typename _Tp, int m, int n> class Matx;
+
+template<typename _Tp> class Complex;
+template<typename _Tp> class Point_;
+template<typename _Tp> class Point3_;
+template<typename _Tp> class Size_;
+template<typename _Tp> class Rect_;
+template<typename _Tp> class Scalar_;
+
+class CV_EXPORTS RotatedRect;
+class CV_EXPORTS Range;
+class CV_EXPORTS TermCriteria;
+class CV_EXPORTS KeyPoint;
+class CV_EXPORTS DMatch;
+class CV_EXPORTS RNG;
+
+class CV_EXPORTS Mat;
+class CV_EXPORTS MatExpr;
+
+class CV_EXPORTS UMat;
+
+class CV_EXPORTS SparseMat;
+typedef Mat MatND;
+
+template<typename _Tp> class Mat_;
+template<typename _Tp> class SparseMat_;
+
+class CV_EXPORTS MatConstIterator;
+class CV_EXPORTS SparseMatIterator;
+class CV_EXPORTS SparseMatConstIterator;
+template<typename _Tp> class MatIterator_;
+template<typename _Tp> class MatConstIterator_;
+template<typename _Tp> class SparseMatIterator_;
+template<typename _Tp> class SparseMatConstIterator_;
+
+namespace ogl
+{
+    class CV_EXPORTS Buffer;
+    class CV_EXPORTS Texture2D;
+    class CV_EXPORTS Arrays;
+}
+
+namespace cuda
+{
+    class CV_EXPORTS GpuMat;
+    class CV_EXPORTS HostMem;
+    class CV_EXPORTS Stream;
+    class CV_EXPORTS Event;
+}
+
+namespace cudev
+{
+    template <typename _Tp> class GpuMat_;
+}
+
+namespace ipp
+{
+CV_EXPORTS   unsigned long long getIppFeatures();
+CV_EXPORTS   void setIppStatus(int status, const char * const funcname = NULL, const char * const filename = NULL,
+                             int line = 0);
+CV_EXPORTS   int getIppStatus();
+CV_EXPORTS   String getIppErrorLocation();
+CV_EXPORTS_W bool   useIPP();
+CV_EXPORTS_W void   setUseIPP(bool flag);
+CV_EXPORTS_W String getIppVersion();
+
+// IPP Not-Exact mode. This function may force use of IPP then both IPP and OpenCV provide proper results
+// but have internal accuracy differences which have too much direct or indirect impact on accuracy tests.
+CV_EXPORTS_W bool useIPP_NotExact();
+CV_EXPORTS_W void setUseIPP_NotExact(bool flag);
+#ifndef DISABLE_OPENCV_3_COMPATIBILITY
+static inline bool useIPP_NE() { return useIPP_NotExact(); }
+static inline void setUseIPP_NE(bool flag) { setUseIPP_NotExact(flag); }
+#endif
+
+} // ipp
+
+//! @endcond
+
+//! @} core_utils
+
+
+
+
+} // cv
+
+#include "opencv2/core/neon_utils.hpp"
+#include "opencv2/core/vsx_utils.hpp"
+#include "opencv2/core/check.hpp"
+
+#endif //OPENCV_CORE_BASE_HPP

model_zoo/official/lite/app/src/main/cpp/include/opencv2/core/bindings_utils.hpp

@@ -0,0 +1,87 @@
+// This file is part of OpenCV project.
+// It is subject to the license terms in the LICENSE file found in the top-level directory
+// of this distribution and at http://opencv.org/license.html.
+
+#ifndef OPENCV_CORE_BINDINGS_UTILS_HPP
+#define OPENCV_CORE_BINDINGS_UTILS_HPP
+
+#include <opencv2/core/async.hpp>
+#include <opencv2/core/detail/async_promise.hpp>
+
+namespace cv { namespace utils {
+//! @addtogroup core_utils
+//! @{
+
+CV_EXPORTS_W String dumpInputArray(InputArray argument);
+
+CV_EXPORTS_W String dumpInputArrayOfArrays(InputArrayOfArrays argument);
+
+CV_EXPORTS_W String dumpInputOutputArray(InputOutputArray argument);
+
+CV_EXPORTS_W String dumpInputOutputArrayOfArrays(InputOutputArrayOfArrays argument);
+
+CV_WRAP static inline
+String dumpBool(bool argument)
+{
+    return (argument) ? String("Bool: True") : String("Bool: False");
+}
+
+CV_WRAP static inline
+String dumpInt(int argument)
+{
+    return cv::format("Int: %d", argument);
+}
+
+CV_WRAP static inline
+String dumpSizeT(size_t argument)
+{
+    std::ostringstream oss("size_t: ", std::ios::ate);
+    oss << argument;
+    return oss.str();
+}
+
+CV_WRAP static inline
+String dumpFloat(float argument)
+{
+    return cv::format("Float: %.2f", argument);
+}
+
+CV_WRAP static inline
+String dumpDouble(double argument)
+{
+    return cv::format("Double: %.2f", argument);
+}
+
+CV_WRAP static inline
+String dumpCString(const char* argument)
+{
+    return cv::format("String: %s", argument);
+}
+
+CV_WRAP static inline
+AsyncArray testAsyncArray(InputArray argument)
+{
+    AsyncPromise p;
+    p.setValue(argument);
+    return p.getArrayResult();
+}
+
+CV_WRAP static inline
+AsyncArray testAsyncException()
+{
+    AsyncPromise p;
+    try
+    {
+        CV_Error(Error::StsOk, "Test: Generated async error");
+    }
+    catch (const cv::Exception& e)
+    {
+        p.setException(e);
+    }
+    return p.getArrayResult();
+}
+
+//! @}
+}} // namespace
+
+#endif // OPENCV_CORE_BINDINGS_UTILS_HPP

model_zoo/official/lite/app/src/main/cpp/include/opencv2/core/bufferpool.hpp

@@ -0,0 +1,40 @@
+// This file is part of OpenCV project.
+// It is subject to the license terms in the LICENSE file found in the top-level directory
+// of this distribution and at http://opencv.org/license.html.
+//
+// Copyright (C) 2014, Advanced Micro Devices, Inc., all rights reserved.
+
+#ifndef OPENCV_CORE_BUFFER_POOL_HPP
+#define OPENCV_CORE_BUFFER_POOL_HPP
+
+#ifdef _MSC_VER
+#pragma warning(push)
+#pragma warning(disable: 4265)
+#endif
+
+namespace cv
+{
+
+//! @addtogroup core
+//! @{
+
+class BufferPoolController
+{
+protected:
+    ~BufferPoolController() { }
+public:
+    virtual size_t getReservedSize() const = 0;
+    virtual size_t getMaxReservedSize() const = 0;
+    virtual void setMaxReservedSize(size_t size) = 0;
+    virtual void freeAllReservedBuffers() = 0;
+};
+
+//! @}
+
+}
+
+#ifdef _MSC_VER
+#pragma warning(pop)
+#endif
+
+#endif // OPENCV_CORE_BUFFER_POOL_HPP

model_zoo/official/lite/app/src/main/cpp/include/opencv2/core/check.hpp

@@ -0,0 +1,160 @@
+// This file is part of OpenCV project.
+// It is subject to the license terms in the LICENSE file found in the top-level directory
+// of this distribution and at http://opencv.org/license.html.
+
+#ifndef OPENCV_CORE_CHECK_HPP
+#define OPENCV_CORE_CHECK_HPP
+
+#include <opencv2/core/base.hpp>
+
+namespace cv {
+
+/** Returns string of cv::Mat depth value: CV_8U -> "CV_8U" or "<invalid depth>" */
+CV_EXPORTS const char* depthToString(int depth);
+
+/** Returns string of cv::Mat depth value: CV_8UC3 -> "CV_8UC3" or "<invalid type>" */
+CV_EXPORTS const String typeToString(int type);
+
+
+//! @cond IGNORED
+namespace detail {
+
+/** Returns string of cv::Mat depth value: CV_8U -> "CV_8U" or NULL */
+CV_EXPORTS const char* depthToString_(int depth);
+
+/** Returns string of cv::Mat depth value: CV_8UC3 -> "CV_8UC3" or cv::String() */
+CV_EXPORTS const cv::String typeToString_(int type);
+
+enum TestOp {
+  TEST_CUSTOM = 0,
+  TEST_EQ = 1,
+  TEST_NE = 2,
+  TEST_LE = 3,
+  TEST_LT = 4,
+  TEST_GE = 5,
+  TEST_GT = 6,
+  CV__LAST_TEST_OP
+};
+
+struct CheckContext {
+    const char* func;
+    const char* file;
+    int line;
+    enum TestOp testOp;
+    const char* message;
+    const char* p1_str;
+    const char* p2_str;
+};
+
+#ifndef CV__CHECK_FILENAME
+# define CV__CHECK_FILENAME __FILE__
+#endif
+
+#ifndef CV__CHECK_FUNCTION
+# if defined _MSC_VER
+#   define CV__CHECK_FUNCTION __FUNCSIG__
+# elif defined __GNUC__
+#   define CV__CHECK_FUNCTION __PRETTY_FUNCTION__
+# else
+#   define CV__CHECK_FUNCTION "<unknown>"
+# endif
+#endif
+
+#define CV__CHECK_LOCATION_VARNAME(id) CVAUX_CONCAT(CVAUX_CONCAT(__cv_check_, id), __LINE__)
+#define CV__DEFINE_CHECK_CONTEXT(id, message, testOp, p1_str, p2_str) \
+    static const cv::detail::CheckContext CV__CHECK_LOCATION_VARNAME(id) = \
+            { CV__CHECK_FUNCTION, CV__CHECK_FILENAME, __LINE__, testOp, message, p1_str, p2_str }
+
+CV_EXPORTS void CV_NORETURN check_failed_auto(const int v1, const int v2, const CheckContext& ctx);
+CV_EXPORTS void CV_NORETURN check_failed_auto(const size_t v1, const size_t v2, const CheckContext& ctx);
+CV_EXPORTS void CV_NORETURN check_failed_auto(const float v1, const float v2, const CheckContext& ctx);
+CV_EXPORTS void CV_NORETURN check_failed_auto(const double v1, const double v2, const CheckContext& ctx);
+CV_EXPORTS void CV_NORETURN check_failed_auto(const Size_<int> v1, const Size_<int> v2, const CheckContext& ctx);
+CV_EXPORTS void CV_NORETURN check_failed_MatDepth(const int v1, const int v2, const CheckContext& ctx);
+CV_EXPORTS void CV_NORETURN check_failed_MatType(const int v1, const int v2, const CheckContext& ctx);
+CV_EXPORTS void CV_NORETURN check_failed_MatChannels(const int v1, const int v2, const CheckContext& ctx);
+
+CV_EXPORTS void CV_NORETURN check_failed_auto(const int v, const CheckContext& ctx);
+CV_EXPORTS void CV_NORETURN check_failed_auto(const size_t v, const CheckContext& ctx);
+CV_EXPORTS void CV_NORETURN check_failed_auto(const float v, const CheckContext& ctx);
+CV_EXPORTS void CV_NORETURN check_failed_auto(const double v, const CheckContext& ctx);
+CV_EXPORTS void CV_NORETURN check_failed_auto(const Size_<int> v, const CheckContext& ctx);
+CV_EXPORTS void CV_NORETURN check_failed_auto(const std::string& v1, const CheckContext& ctx);
+CV_EXPORTS void CV_NORETURN check_failed_MatDepth(const int v, const CheckContext& ctx);
+CV_EXPORTS void CV_NORETURN check_failed_MatType(const int v, const CheckContext& ctx);
+CV_EXPORTS void CV_NORETURN check_failed_MatChannels(const int v, const CheckContext& ctx);
+
+
+#define CV__TEST_EQ(v1, v2) ((v1) == (v2))
+#define CV__TEST_NE(v1, v2) ((v1) != (v2))
+#define CV__TEST_LE(v1, v2) ((v1) <= (v2))
+#define CV__TEST_LT(v1, v2) ((v1) < (v2))
+#define CV__TEST_GE(v1, v2) ((v1) >= (v2))
+#define CV__TEST_GT(v1, v2) ((v1) > (v2))
+
+#define CV__CHECK(id, op, type, v1, v2, v1_str, v2_str, msg_str) do { \
+    if(CV__TEST_##op((v1), (v2))) ; else { \
+        CV__DEFINE_CHECK_CONTEXT(id, msg_str, cv::detail::TEST_ ## op, v1_str, v2_str); \
+        cv::detail::check_failed_ ## type((v1), (v2), CV__CHECK_LOCATION_VARNAME(id)); \
+    } \
+} while (0)
+
+#define CV__CHECK_CUSTOM_TEST(id, type, v, test_expr, v_str, test_expr_str, msg_str) do { \
+    if(!!(test_expr)) ; else { \
+        CV__DEFINE_CHECK_CONTEXT(id, msg_str, cv::detail::TEST_CUSTOM, v_str, test_expr_str); \
+        cv::detail::check_failed_ ## type((v), CV__CHECK_LOCATION_VARNAME(id)); \
+    } \
+} while (0)
+
+} // namespace
+//! @endcond
+
+
+/// Supported values of these types: int, float, double
+#define CV_CheckEQ(v1, v2, msg)  CV__CHECK(_, EQ, auto, v1, v2, #v1, #v2, msg)
+#define CV_CheckNE(v1, v2, msg)  CV__CHECK(_, NE, auto, v1, v2, #v1, #v2, msg)
+#define CV_CheckLE(v1, v2, msg)  CV__CHECK(_, LE, auto, v1, v2, #v1, #v2, msg)
+#define CV_CheckLT(v1, v2, msg)  CV__CHECK(_, LT, auto, v1, v2, #v1, #v2, msg)
+#define CV_CheckGE(v1, v2, msg)  CV__CHECK(_, GE, auto, v1, v2, #v1, #v2, msg)
+#define CV_CheckGT(v1, v2, msg)  CV__CHECK(_, GT, auto, v1, v2, #v1, #v2, msg)
+
+/// Check with additional "decoding" of type values in error message
+#define CV_CheckTypeEQ(t1, t2, msg)  CV__CHECK(_, EQ, MatType, t1, t2, #t1, #t2, msg)
+/// Check with additional "decoding" of depth values in error message
+#define CV_CheckDepthEQ(d1, d2, msg)  CV__CHECK(_, EQ, MatDepth, d1, d2, #d1, #d2, msg)
+
+#define CV_CheckChannelsEQ(c1, c2, msg)  CV__CHECK(_, EQ, MatChannels, c1, c2, #c1, #c2, msg)
+
+/// Example: type == CV_8UC1 || type == CV_8UC3
+#define CV_CheckType(t, test_expr, msg)  CV__CHECK_CUSTOM_TEST(_, MatType, t, (test_expr), #t, #test_expr, msg)
+
+/// Example: depth == CV_32F || depth == CV_64F
+#define CV_CheckDepth(t, test_expr, msg)  CV__CHECK_CUSTOM_TEST(_, MatDepth, t, (test_expr), #t, #test_expr, msg)
+
+/// Example: v == A || v == B
+#define CV_Check(v, test_expr, msg)  CV__CHECK_CUSTOM_TEST(_, auto, v, (test_expr), #v, #test_expr, msg)
+
+/// Some complex conditions: CV_Check(src2, src2.empty() || (src2.type() == src1.type() && src2.size() == src1.size()), "src2 should have same size/type as src1")
+// TODO define pretty-printers
+
+#ifndef NDEBUG
+#define CV_DbgCheck(v, test_expr, msg)  CV__CHECK_CUSTOM_TEST(_, auto, v, (test_expr), #v, #test_expr, msg)
+#define CV_DbgCheckEQ(v1, v2, msg)  CV__CHECK(_, EQ, auto, v1, v2, #v1, #v2, msg)
+#define CV_DbgCheckNE(v1, v2, msg)  CV__CHECK(_, NE, auto, v1, v2, #v1, #v2, msg)
+#define CV_DbgCheckLE(v1, v2, msg)  CV__CHECK(_, LE, auto, v1, v2, #v1, #v2, msg)
+#define CV_DbgCheckLT(v1, v2, msg)  CV__CHECK(_, LT, auto, v1, v2, #v1, #v2, msg)
+#define CV_DbgCheckGE(v1, v2, msg)  CV__CHECK(_, GE, auto, v1, v2, #v1, #v2, msg)
+#define CV_DbgCheckGT(v1, v2, msg)  CV__CHECK(_, GT, auto, v1, v2, #v1, #v2, msg)
+#else
+#define CV_DbgCheck(v, test_expr, msg)  do { } while (0)
+#define CV_DbgCheckEQ(v1, v2, msg)  do { } while (0)
+#define CV_DbgCheckNE(v1, v2, msg)  do { } while (0)
+#define CV_DbgCheckLE(v1, v2, msg)  do { } while (0)
+#define CV_DbgCheckLT(v1, v2, msg)  do { } while (0)
+#define CV_DbgCheckGE(v1, v2, msg)  do { } while (0)
+#define CV_DbgCheckGT(v1, v2, msg)  do { } while (0)
+#endif
+
+} // namespace
+
+#endif // OPENCV_CORE_CHECK_HPP

model_zoo/official/lite/app/src/main/cpp/include/opencv2/core/core.hpp

@@ -0,0 +1,48 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifdef __OPENCV_BUILD
+#error this is a compatibility header which should not be used inside the OpenCV library
+#endif
+
+#include "opencv2/core.hpp"

model_zoo/official/lite/app/src/main/cpp/include/opencv2/core/cuda.inl.hpp

@@ -0,0 +1,637 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CORE_CUDAINL_HPP
+#define OPENCV_CORE_CUDAINL_HPP
+
+#include "opencv2/core/cuda.hpp"
+
+//! @cond IGNORED
+
+namespace cv { namespace cuda {
+
+//===================================================================================
+// GpuMat
+//===================================================================================
+
+inline
+GpuMat::GpuMat(Allocator* allocator_)
+    : flags(0), rows(0), cols(0), step(0), data(0), refcount(0), datastart(0), dataend(0), allocator(allocator_)
+{}
+
+inline
+GpuMat::GpuMat(int rows_, int cols_, int type_, Allocator* allocator_)
+    : flags(0), rows(0), cols(0), step(0), data(0), refcount(0), datastart(0), dataend(0), allocator(allocator_)
+{
+    if (rows_ > 0 && cols_ > 0)
+        create(rows_, cols_, type_);
+}
+
+inline
+GpuMat::GpuMat(Size size_, int type_, Allocator* allocator_)
+    : flags(0), rows(0), cols(0), step(0), data(0), refcount(0), datastart(0), dataend(0), allocator(allocator_)
+{
+    if (size_.height > 0 && size_.width > 0)
+        create(size_.height, size_.width, type_);
+}
+
+inline
+GpuMat::GpuMat(int rows_, int cols_, int type_, Scalar s_, Allocator* allocator_)
+    : flags(0), rows(0), cols(0), step(0), data(0), refcount(0), datastart(0), dataend(0), allocator(allocator_)
+{
+    if (rows_ > 0 && cols_ > 0)
+    {
+        create(rows_, cols_, type_);
+        setTo(s_);
+    }
+}
+
+inline
+GpuMat::GpuMat(Size size_, int type_, Scalar s_, Allocator* allocator_)
+    : flags(0), rows(0), cols(0), step(0), data(0), refcount(0), datastart(0), dataend(0), allocator(allocator_)
+{
+    if (size_.height > 0 && size_.width > 0)
+    {
+        create(size_.height, size_.width, type_);
+        setTo(s_);
+    }
+}
+
+inline
+GpuMat::GpuMat(const GpuMat& m)
+    : flags(m.flags), rows(m.rows), cols(m.cols), step(m.step), data(m.data), refcount(m.refcount), datastart(m.datastart), dataend(m.dataend), allocator(m.allocator)
+{
+    if (refcount)
+        CV_XADD(refcount, 1);
+}
+
+inline
+GpuMat::GpuMat(InputArray arr, Allocator* allocator_) :
+    flags(0), rows(0), cols(0), step(0), data(0), refcount(0), datastart(0), dataend(0), allocator(allocator_)
+{
+    upload(arr);
+}
+
+inline
+GpuMat::~GpuMat()
+{
+    release();
+}
+
+inline
+GpuMat& GpuMat::operator =(const GpuMat& m)
+{
+    if (this != &m)
+    {
+        GpuMat temp(m);
+        swap(temp);
+    }
+
+    return *this;
+}
+
+inline
+void GpuMat::create(Size size_, int type_)
+{
+    create(size_.height, size_.width, type_);
+}
+
+inline
+void GpuMat::swap(GpuMat& b)
+{
+    std::swap(flags, b.flags);
+    std::swap(rows, b.rows);
+    std::swap(cols, b.cols);
+    std::swap(step, b.step);
+    std::swap(data, b.data);
+    std::swap(datastart, b.datastart);
+    std::swap(dataend, b.dataend);
+    std::swap(refcount, b.refcount);
+    std::swap(allocator, b.allocator);
+}
+
+inline
+GpuMat GpuMat::clone() const
+{
+    GpuMat m;
+    copyTo(m);
+    return m;
+}
+
+inline
+void GpuMat::copyTo(OutputArray dst, InputArray mask) const
+{
+    copyTo(dst, mask, Stream::Null());
+}
+
+inline
+GpuMat& GpuMat::setTo(Scalar s)
+{
+    return setTo(s, Stream::Null());
+}
+
+inline
+GpuMat& GpuMat::setTo(Scalar s, InputArray mask)
+{
+    return setTo(s, mask, Stream::Null());
+}
+
+inline
+void GpuMat::convertTo(OutputArray dst, int rtype) const
+{
+    convertTo(dst, rtype, Stream::Null());
+}
+
+inline
+void GpuMat::convertTo(OutputArray dst, int rtype, double alpha, double beta) const
+{
+    convertTo(dst, rtype, alpha, beta, Stream::Null());
+}
+
+inline
+void GpuMat::convertTo(OutputArray dst, int rtype, double alpha, Stream& stream) const
+{
+    convertTo(dst, rtype, alpha, 0.0, stream);
+}
+
+inline
+void GpuMat::assignTo(GpuMat& m, int _type) const
+{
+    if (_type < 0)
+        m = *this;
+    else
+        convertTo(m, _type);
+}
+
+inline
+uchar* GpuMat::ptr(int y)
+{
+    CV_DbgAssert( (unsigned)y < (unsigned)rows );
+    return data + step * y;
+}
+
+inline
+const uchar* GpuMat::ptr(int y) const
+{
+    CV_DbgAssert( (unsigned)y < (unsigned)rows );
+    return data + step * y;
+}
+
+template<typename _Tp> inline
+_Tp* GpuMat::ptr(int y)
+{
+    return (_Tp*)ptr(y);
+}
+
+template<typename _Tp> inline
+const _Tp* GpuMat::ptr(int y) const
+{
+    return (const _Tp*)ptr(y);
+}
+
+template <class T> inline
+GpuMat::operator PtrStepSz<T>() const
+{
+    return PtrStepSz<T>(rows, cols, (T*)data, step);
+}
+
+template <class T> inline
+GpuMat::operator PtrStep<T>() const
+{
+    return PtrStep<T>((T*)data, step);
+}
+
+inline
+GpuMat GpuMat::row(int y) const
+{
+    return GpuMat(*this, Range(y, y+1), Range::all());
+}
+
+inline
+GpuMat GpuMat::col(int x) const
+{
+    return GpuMat(*this, Range::all(), Range(x, x+1));
+}
+
+inline
+GpuMat GpuMat::rowRange(int startrow, int endrow) const
+{
+    return GpuMat(*this, Range(startrow, endrow), Range::all());
+}
+
+inline
+GpuMat GpuMat::rowRange(Range r) const
+{
+    return GpuMat(*this, r, Range::all());
+}
+
+inline
+GpuMat GpuMat::colRange(int startcol, int endcol) const
+{
+    return GpuMat(*this, Range::all(), Range(startcol, endcol));
+}
+
+inline
+GpuMat GpuMat::colRange(Range r) const
+{
+    return GpuMat(*this, Range::all(), r);
+}
+
+inline
+GpuMat GpuMat::operator ()(Range rowRange_, Range colRange_) const
+{
+    return GpuMat(*this, rowRange_, colRange_);
+}
+
+inline
+GpuMat GpuMat::operator ()(Rect roi) const
+{
+    return GpuMat(*this, roi);
+}
+
+inline
+bool GpuMat::isContinuous() const
+{
+    return (flags & Mat::CONTINUOUS_FLAG) != 0;
+}
+
+inline
+size_t GpuMat::elemSize() const
+{
+    return CV_ELEM_SIZE(flags);
+}
+
+inline
+size_t GpuMat::elemSize1() const
+{
+    return CV_ELEM_SIZE1(flags);
+}
+
+inline
+int GpuMat::type() const
+{
+    return CV_MAT_TYPE(flags);
+}
+
+inline
+int GpuMat::depth() const
+{
+    return CV_MAT_DEPTH(flags);
+}
+
+inline
+int GpuMat::channels() const
+{
+    return CV_MAT_CN(flags);
+}
+
+inline
+size_t GpuMat::step1() const
+{
+    return step / elemSize1();
+}
+
+inline
+Size GpuMat::size() const
+{
+    return Size(cols, rows);
+}
+
+inline
+bool GpuMat::empty() const
+{
+    return data == 0;
+}
+
+inline
+void* GpuMat::cudaPtr() const
+{
+    return data;
+}
+
+static inline
+GpuMat createContinuous(int rows, int cols, int type)
+{
+    GpuMat m;
+    createContinuous(rows, cols, type, m);
+    return m;
+}
+
+static inline
+void createContinuous(Size size, int type, OutputArray arr)
+{
+    createContinuous(size.height, size.width, type, arr);
+}
+
+static inline
+GpuMat createContinuous(Size size, int type)
+{
+    GpuMat m;
+    createContinuous(size, type, m);
+    return m;
+}
+
+static inline
+void ensureSizeIsEnough(Size size, int type, OutputArray arr)
+{
+    ensureSizeIsEnough(size.height, size.width, type, arr);
+}
+
+static inline
+void swap(GpuMat& a, GpuMat& b)
+{
+    a.swap(b);
+}
+
+//===================================================================================
+// HostMem
+//===================================================================================
+
+inline
+HostMem::HostMem(AllocType alloc_type_)
+    : flags(0), rows(0), cols(0), step(0), data(0), refcount(0), datastart(0), dataend(0), alloc_type(alloc_type_)
+{
+}
+
+inline
+HostMem::HostMem(const HostMem& m)
+    : flags(m.flags), rows(m.rows), cols(m.cols), step(m.step), data(m.data), refcount(m.refcount), datastart(m.datastart), dataend(m.dataend), alloc_type(m.alloc_type)
+{
+    if( refcount )
+        CV_XADD(refcount, 1);
+}
+
+inline
+HostMem::HostMem(int rows_, int cols_, int type_, AllocType alloc_type_)
+    : flags(0), rows(0), cols(0), step(0), data(0), refcount(0), datastart(0), dataend(0), alloc_type(alloc_type_)
+{
+    if (rows_ > 0 && cols_ > 0)
+        create(rows_, cols_, type_);
+}
+
+inline
+HostMem::HostMem(Size size_, int type_, AllocType alloc_type_)
+    : flags(0), rows(0), cols(0), step(0), data(0), refcount(0), datastart(0), dataend(0), alloc_type(alloc_type_)
+{
+    if (size_.height > 0 && size_.width > 0)
+        create(size_.height, size_.width, type_);
+}
+
+inline
+HostMem::HostMem(InputArray arr, AllocType alloc_type_)
+    : flags(0), rows(0), cols(0), step(0), data(0), refcount(0), datastart(0), dataend(0), alloc_type(alloc_type_)
+{
+    arr.getMat().copyTo(*this);
+}
+
+inline
+HostMem::~HostMem()
+{
+    release();
+}
+
+inline
+HostMem& HostMem::operator =(const HostMem& m)
+{
+    if (this != &m)
+    {
+        HostMem temp(m);
+        swap(temp);
+    }
+
+    return *this;
+}
+
+inline
+void HostMem::swap(HostMem& b)
+{
+    std::swap(flags, b.flags);
+    std::swap(rows, b.rows);
+    std::swap(cols, b.cols);
+    std::swap(step, b.step);
+    std::swap(data, b.data);
+    std::swap(datastart, b.datastart);
+    std::swap(dataend, b.dataend);
+    std::swap(refcount, b.refcount);
+    std::swap(alloc_type, b.alloc_type);
+}
+
+inline
+HostMem HostMem::clone() const
+{
+    HostMem m(size(), type(), alloc_type);
+    createMatHeader().copyTo(m);
+    return m;
+}
+
+inline
+void HostMem::create(Size size_, int type_)
+{
+    create(size_.height, size_.width, type_);
+}
+
+inline
+Mat HostMem::createMatHeader() const
+{
+    return Mat(size(), type(), data, step);
+}
+
+inline
+bool HostMem::isContinuous() const
+{
+    return (flags & Mat::CONTINUOUS_FLAG) != 0;
+}
+
+inline
+size_t HostMem::elemSize() const
+{
+    return CV_ELEM_SIZE(flags);
+}
+
+inline
+size_t HostMem::elemSize1() const
+{
+    return CV_ELEM_SIZE1(flags);
+}
+
+inline
+int HostMem::type() const
+{
+    return CV_MAT_TYPE(flags);
+}
+
+inline
+int HostMem::depth() const
+{
+    return CV_MAT_DEPTH(flags);
+}
+
+inline
+int HostMem::channels() const
+{
+    return CV_MAT_CN(flags);
+}
+
+inline
+size_t HostMem::step1() const
+{
+    return step / elemSize1();
+}
+
+inline
+Size HostMem::size() const
+{
+    return Size(cols, rows);
+}
+
+inline
+bool HostMem::empty() const
+{
+    return data == 0;
+}
+
+static inline
+void swap(HostMem& a, HostMem& b)
+{
+    a.swap(b);
+}
+
+//===================================================================================
+// Stream
+//===================================================================================
+
+inline
+Stream::Stream(const Ptr<Impl>& impl)
+    : impl_(impl)
+{
+}
+
+//===================================================================================
+// Event
+//===================================================================================
+
+inline
+Event::Event(const Ptr<Impl>& impl)
+    : impl_(impl)
+{
+}
+
+//===================================================================================
+// Initialization & Info
+//===================================================================================
+
+inline
+bool TargetArchs::has(int major, int minor)
+{
+    return hasPtx(major, minor) || hasBin(major, minor);
+}
+
+inline
+bool TargetArchs::hasEqualOrGreater(int major, int minor)
+{
+    return hasEqualOrGreaterPtx(major, minor) || hasEqualOrGreaterBin(major, minor);
+}
+
+inline
+DeviceInfo::DeviceInfo()
+{
+    device_id_ = getDevice();
+}
+
+inline
+DeviceInfo::DeviceInfo(int device_id)
+{
+    CV_Assert( device_id >= 0 && device_id < getCudaEnabledDeviceCount() );
+    device_id_ = device_id;
+}
+
+inline
+int DeviceInfo::deviceID() const
+{
+    return device_id_;
+}
+
+inline
+size_t DeviceInfo::freeMemory() const
+{
+    size_t _totalMemory = 0, _freeMemory = 0;
+    queryMemory(_totalMemory, _freeMemory);
+    return _freeMemory;
+}
+
+inline
+size_t DeviceInfo::totalMemory() const
+{
+    size_t _totalMemory = 0, _freeMemory = 0;
+    queryMemory(_totalMemory, _freeMemory);
+    return _totalMemory;
+}
+
+inline
+bool DeviceInfo::supports(FeatureSet feature_set) const
+{
+    int version = majorVersion() * 10 + minorVersion();
+    return version >= feature_set;
+}
+
+
+}} // namespace cv { namespace cuda {
+
+//===================================================================================
+// Mat
+//===================================================================================
+
+namespace cv {
+
+inline
+Mat::Mat(const cuda::GpuMat& m)
+    : flags(0), dims(0), rows(0), cols(0), data(0), datastart(0), dataend(0), datalimit(0), allocator(0), u(0), size(&rows)
+{
+    m.download(*this);
+}
+
+}
+
+//! @endcond
+
+#endif // OPENCV_CORE_CUDAINL_HPP

model_zoo/official/lite/app/src/main/cpp/include/opencv2/core/cuda/block.hpp

@@ -0,0 +1,211 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CUDA_DEVICE_BLOCK_HPP
+#define OPENCV_CUDA_DEVICE_BLOCK_HPP
+
+/** @file
+ * @deprecated Use @ref cudev instead.
+ */
+
+//! @cond IGNORED
+
+namespace cv { namespace cuda { namespace device
+{
+    struct Block
+    {
+        static __device__ __forceinline__ unsigned int id()
+        {
+            return blockIdx.x;
+        }
+
+        static __device__ __forceinline__ unsigned int stride()
+        {
+            return blockDim.x * blockDim.y * blockDim.z;
+        }
+
+        static __device__ __forceinline__ void sync()
+        {
+            __syncthreads();
+        }
+
+        static __device__ __forceinline__ int flattenedThreadId()
+        {
+            return threadIdx.z * blockDim.x * blockDim.y + threadIdx.y * blockDim.x + threadIdx.x;
+        }
+
+        template<typename It, typename T>
+        static __device__ __forceinline__ void fill(It beg, It end, const T& value)
+        {
+            int STRIDE = stride();
+            It t = beg + flattenedThreadId();
+
+            for(; t < end; t += STRIDE)
+                *t = value;
+        }
+
+        template<typename OutIt, typename T>
+        static __device__ __forceinline__ void yota(OutIt beg, OutIt end, T value)
+        {
+            int STRIDE = stride();
+            int tid = flattenedThreadId();
+            value += tid;
+
+            for(OutIt t = beg + tid; t < end; t += STRIDE, value += STRIDE)
+                *t = value;
+        }
+
+        template<typename InIt, typename OutIt>
+        static __device__ __forceinline__ void copy(InIt beg, InIt end, OutIt out)
+        {
+            int STRIDE = stride();
+            InIt  t = beg + flattenedThreadId();
+            OutIt o = out + (t - beg);
+
+            for(; t < end; t += STRIDE, o += STRIDE)
+                *o = *t;
+        }
+
+        template<typename InIt, typename OutIt, class UnOp>
+        static __device__ __forceinline__ void transform(InIt beg, InIt end, OutIt out, UnOp op)
+        {
+            int STRIDE = stride();
+            InIt  t = beg + flattenedThreadId();
+            OutIt o = out + (t - beg);
+
+            for(; t < end; t += STRIDE, o += STRIDE)
+                *o = op(*t);
+        }
+
+        template<typename InIt1, typename InIt2, typename OutIt, class BinOp>
+        static __device__ __forceinline__ void transform(InIt1 beg1, InIt1 end1, InIt2 beg2, OutIt out, BinOp op)
+        {
+            int STRIDE = stride();
+            InIt1 t1 = beg1 + flattenedThreadId();
+            InIt2 t2 = beg2 + flattenedThreadId();
+            OutIt o  = out + (t1 - beg1);
+
+            for(; t1 < end1; t1 += STRIDE, t2 += STRIDE, o += STRIDE)
+                *o = op(*t1, *t2);
+        }
+
+        template<int CTA_SIZE, typename T, class BinOp>
+        static __device__ __forceinline__ void reduce(volatile T* buffer, BinOp op)
+        {
+            int tid = flattenedThreadId();
+            T val =  buffer[tid];
+
+            if (CTA_SIZE >= 1024) { if (tid < 512) buffer[tid] = val = op(val, buffer[tid + 512]); __syncthreads(); }
+            if (CTA_SIZE >=  512) { if (tid < 256) buffer[tid] = val = op(val, buffer[tid + 256]); __syncthreads(); }
+            if (CTA_SIZE >=  256) { if (tid < 128) buffer[tid] = val = op(val, buffer[tid + 128]); __syncthreads(); }
+            if (CTA_SIZE >=  128) { if (tid <  64) buffer[tid] = val = op(val, buffer[tid +  64]); __syncthreads(); }
+
+            if (tid < 32)
+            {
+                if (CTA_SIZE >=   64) { buffer[tid] = val = op(val, buffer[tid +  32]); }
+                if (CTA_SIZE >=   32) { buffer[tid] = val = op(val, buffer[tid +  16]); }
+                if (CTA_SIZE >=   16) { buffer[tid] = val = op(val, buffer[tid +   8]); }
+                if (CTA_SIZE >=    8) { buffer[tid] = val = op(val, buffer[tid +   4]); }
+                if (CTA_SIZE >=    4) { buffer[tid] = val = op(val, buffer[tid +   2]); }
+                if (CTA_SIZE >=    2) { buffer[tid] = val = op(val, buffer[tid +   1]); }
+            }
+        }
+
+        template<int CTA_SIZE, typename T, class BinOp>
+        static __device__ __forceinline__ T reduce(volatile T* buffer, T init, BinOp op)
+        {
+            int tid = flattenedThreadId();
+            T val =  buffer[tid] = init;
+            __syncthreads();
+
+            if (CTA_SIZE >= 1024) { if (tid < 512) buffer[tid] = val = op(val, buffer[tid + 512]); __syncthreads(); }
+            if (CTA_SIZE >=  512) { if (tid < 256) buffer[tid] = val = op(val, buffer[tid + 256]); __syncthreads(); }
+            if (CTA_SIZE >=  256) { if (tid < 128) buffer[tid] = val = op(val, buffer[tid + 128]); __syncthreads(); }
+            if (CTA_SIZE >=  128) { if (tid <  64) buffer[tid] = val = op(val, buffer[tid +  64]); __syncthreads(); }
+
+            if (tid < 32)
+            {
+                if (CTA_SIZE >=   64) { buffer[tid] = val = op(val, buffer[tid +  32]); }
+                if (CTA_SIZE >=   32) { buffer[tid] = val = op(val, buffer[tid +  16]); }
+                if (CTA_SIZE >=   16) { buffer[tid] = val = op(val, buffer[tid +   8]); }
+                if (CTA_SIZE >=    8) { buffer[tid] = val = op(val, buffer[tid +   4]); }
+                if (CTA_SIZE >=    4) { buffer[tid] = val = op(val, buffer[tid +   2]); }
+                if (CTA_SIZE >=    2) { buffer[tid] = val = op(val, buffer[tid +   1]); }
+            }
+            __syncthreads();
+            return buffer[0];
+        }
+
+        template <typename T, class BinOp>
+        static __device__ __forceinline__ void reduce_n(T* data, unsigned int n, BinOp op)
+        {
+            int ftid = flattenedThreadId();
+            int sft = stride();
+
+            if (sft < n)
+            {
+                for (unsigned int i = sft + ftid; i < n; i += sft)
+                    data[ftid] = op(data[ftid], data[i]);
+
+                __syncthreads();
+
+                n = sft;
+            }
+
+            while (n > 1)
+            {
+                unsigned int half = n/2;
+
+                if (ftid < half)
+                    data[ftid] = op(data[ftid], data[n - ftid - 1]);
+
+                __syncthreads();
+
+                n = n - half;
+            }
+        }
+    };
+}}}
+
+//! @endcond
+
+#endif /* OPENCV_CUDA_DEVICE_BLOCK_HPP */

model_zoo/official/lite/app/src/main/cpp/include/opencv2/core/cuda/border_interpolate.hpp

@@ -0,0 +1,722 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CUDA_BORDER_INTERPOLATE_HPP
+#define OPENCV_CUDA_BORDER_INTERPOLATE_HPP
+
+#include "saturate_cast.hpp"
+#include "vec_traits.hpp"
+#include "vec_math.hpp"
+
+/** @file
+ * @deprecated Use @ref cudev instead.
+ */
+
+//! @cond IGNORED
+
+namespace cv { namespace cuda { namespace device
+{
+    //////////////////////////////////////////////////////////////
+    // BrdConstant
+
+    template <typename D> struct BrdRowConstant
+    {
+        typedef D result_type;
+
+        explicit __host__ __device__ __forceinline__ BrdRowConstant(int width_, const D& val_ = VecTraits<D>::all(0)) : width(width_), val(val_) {}
+
+        template <typename T> __device__ __forceinline__ D at_low(int x, const T* data) const
+        {
+            return x >= 0 ? saturate_cast<D>(data[x]) : val;
+        }
+
+        template <typename T> __device__ __forceinline__ D at_high(int x, const T* data) const
+        {
+            return x < width ? saturate_cast<D>(data[x]) : val;
+        }
+
+        template <typename T> __device__ __forceinline__ D at(int x, const T* data) const
+        {
+            return (x >= 0 && x < width) ? saturate_cast<D>(data[x]) : val;
+        }
+
+        int width;
+        D val;
+    };
+
+    template <typename D> struct BrdColConstant
+    {
+        typedef D result_type;
+
+        explicit __host__ __device__ __forceinline__ BrdColConstant(int height_, const D& val_ = VecTraits<D>::all(0)) : height(height_), val(val_) {}
+
+        template <typename T> __device__ __forceinline__ D at_low(int y, const T* data, size_t step) const
+        {
+            return y >= 0 ? saturate_cast<D>(*(const T*)((const char*)data + y * step)) : val;
+        }
+
+        template <typename T> __device__ __forceinline__ D at_high(int y, const T* data, size_t step) const
+        {
+            return y < height ? saturate_cast<D>(*(const T*)((const char*)data + y * step)) : val;
+        }
+
+        template <typename T> __device__ __forceinline__ D at(int y, const T* data, size_t step) const
+        {
+            return (y >= 0 && y < height) ? saturate_cast<D>(*(const T*)((const char*)data + y * step)) : val;
+        }
+
+        int height;
+        D val;
+    };
+
+    template <typename D> struct BrdConstant
+    {
+        typedef D result_type;
+
+        __host__ __device__ __forceinline__ BrdConstant(int height_, int width_, const D& val_ = VecTraits<D>::all(0)) : height(height_), width(width_), val(val_)
+        {
+        }
+
+        template <typename T> __device__ __forceinline__ D at(int y, int x, const T* data, size_t step) const
+        {
+            return (x >= 0 && x < width && y >= 0 && y < height) ? saturate_cast<D>(((const T*)((const uchar*)data + y * step))[x]) : val;
+        }
+
+        template <typename Ptr2D> __device__ __forceinline__ D at(typename Ptr2D::index_type y, typename Ptr2D::index_type x, const Ptr2D& src) const
+        {
+            return (x >= 0 && x < width && y >= 0 && y < height) ? saturate_cast<D>(src(y, x)) : val;
+        }
+
+        int height;
+        int width;
+        D val;
+    };
+
+    //////////////////////////////////////////////////////////////
+    // BrdReplicate
+
+    template <typename D> struct BrdRowReplicate
+    {
+        typedef D result_type;
+
+        explicit __host__ __device__ __forceinline__ BrdRowReplicate(int width) : last_col(width - 1) {}
+        template <typename U> __host__ __device__ __forceinline__ BrdRowReplicate(int width, U) : last_col(width - 1) {}
+
+        __device__ __forceinline__ int idx_col_low(int x) const
+        {
+            return ::max(x, 0);
+        }
+
+        __device__ __forceinline__ int idx_col_high(int x) const
+        {
+            return ::min(x, last_col);
+        }
+
+        __device__ __forceinline__ int idx_col(int x) const
+        {
+            return idx_col_low(idx_col_high(x));
+        }
+
+        template <typename T> __device__ __forceinline__ D at_low(int x, const T* data) const
+        {
+            return saturate_cast<D>(data[idx_col_low(x)]);
+        }
+
+        template <typename T> __device__ __forceinline__ D at_high(int x, const T* data) const
+        {
+            return saturate_cast<D>(data[idx_col_high(x)]);
+        }
+
+        template <typename T> __device__ __forceinline__ D at(int x, const T* data) const
+        {
+            return saturate_cast<D>(data[idx_col(x)]);
+        }
+
+        int last_col;
+    };
+
+    template <typename D> struct BrdColReplicate
+    {
+        typedef D result_type;
+
+        explicit __host__ __device__ __forceinline__ BrdColReplicate(int height) : last_row(height - 1) {}
+        template <typename U> __host__ __device__ __forceinline__ BrdColReplicate(int height, U) : last_row(height - 1) {}
+
+        __device__ __forceinline__ int idx_row_low(int y) const
+        {
+            return ::max(y, 0);
+        }
+
+        __device__ __forceinline__ int idx_row_high(int y) const
+        {
+            return ::min(y, last_row);
+        }
+
+        __device__ __forceinline__ int idx_row(int y) const
+        {
+            return idx_row_low(idx_row_high(y));
+        }
+
+        template <typename T> __device__ __forceinline__ D at_low(int y, const T* data, size_t step) const
+        {
+            return saturate_cast<D>(*(const T*)((const char*)data + idx_row_low(y) * step));
+        }
+
+        template <typename T> __device__ __forceinline__ D at_high(int y, const T* data, size_t step) const
+        {
+            return saturate_cast<D>(*(const T*)((const char*)data + idx_row_high(y) * step));
+        }
+
+        template <typename T> __device__ __forceinline__ D at(int y, const T* data, size_t step) const
+        {
+            return saturate_cast<D>(*(const T*)((const char*)data + idx_row(y) * step));
+        }
+
+        int last_row;
+    };
+
+    template <typename D> struct BrdReplicate
+    {
+        typedef D result_type;
+
+        __host__ __device__ __forceinline__ BrdReplicate(int height, int width) : last_row(height - 1), last_col(width - 1) {}
+        template <typename U> __host__ __device__ __forceinline__ BrdReplicate(int height, int width, U) : last_row(height - 1), last_col(width - 1) {}
+
+        __device__ __forceinline__ int idx_row_low(int y) const
+        {
+            return ::max(y, 0);
+        }
+
+        __device__ __forceinline__ int idx_row_high(int y) const
+        {
+            return ::min(y, last_row);
+        }
+
+        __device__ __forceinline__ int idx_row(int y) const
+        {
+            return idx_row_low(idx_row_high(y));
+        }
+
+        __device__ __forceinline__ int idx_col_low(int x) const
+        {
+            return ::max(x, 0);
+        }
+
+        __device__ __forceinline__ int idx_col_high(int x) const
+        {
+            return ::min(x, last_col);
+        }
+
+        __device__ __forceinline__ int idx_col(int x) const
+        {
+            return idx_col_low(idx_col_high(x));
+        }
+
+        template <typename T> __device__ __forceinline__ D at(int y, int x, const T* data, size_t step) const
+        {
+            return saturate_cast<D>(((const T*)((const char*)data + idx_row(y) * step))[idx_col(x)]);
+        }
+
+        template <typename Ptr2D> __device__ __forceinline__ D at(typename Ptr2D::index_type y, typename Ptr2D::index_type x, const Ptr2D& src) const
+        {
+            return saturate_cast<D>(src(idx_row(y), idx_col(x)));
+        }
+
+        int last_row;
+        int last_col;
+    };
+
+    //////////////////////////////////////////////////////////////
+    // BrdReflect101
+
+    template <typename D> struct BrdRowReflect101
+    {
+        typedef D result_type;
+
+        explicit __host__ __device__ __forceinline__ BrdRowReflect101(int width) : last_col(width - 1) {}
+        template <typename U> __host__ __device__ __forceinline__ BrdRowReflect101(int width, U) : last_col(width - 1) {}
+
+        __device__ __forceinline__ int idx_col_low(int x) const
+        {
+            return ::abs(x) % (last_col + 1);
+        }
+
+        __device__ __forceinline__ int idx_col_high(int x) const
+        {
+            return ::abs(last_col - ::abs(last_col - x)) % (last_col + 1);
+        }
+
+        __device__ __forceinline__ int idx_col(int x) const
+        {
+            return idx_col_low(idx_col_high(x));
+        }
+
+        template <typename T> __device__ __forceinline__ D at_low(int x, const T* data) const
+        {
+            return saturate_cast<D>(data[idx_col_low(x)]);
+        }
+
+        template <typename T> __device__ __forceinline__ D at_high(int x, const T* data) const
+        {
+            return saturate_cast<D>(data[idx_col_high(x)]);
+        }
+
+        template <typename T> __device__ __forceinline__ D at(int x, const T* data) const
+        {
+            return saturate_cast<D>(data[idx_col(x)]);
+        }
+
+        int last_col;
+    };
+
+    template <typename D> struct BrdColReflect101
+    {
+        typedef D result_type;
+
+        explicit __host__ __device__ __forceinline__ BrdColReflect101(int height) : last_row(height - 1) {}
+        template <typename U> __host__ __device__ __forceinline__ BrdColReflect101(int height, U) : last_row(height - 1) {}
+
+        __device__ __forceinline__ int idx_row_low(int y) const
+        {
+            return ::abs(y) % (last_row + 1);
+        }
+
+        __device__ __forceinline__ int idx_row_high(int y) const
+        {
+            return ::abs(last_row - ::abs(last_row - y)) % (last_row + 1);
+        }
+
+        __device__ __forceinline__ int idx_row(int y) const
+        {
+            return idx_row_low(idx_row_high(y));
+        }
+
+        template <typename T> __device__ __forceinline__ D at_low(int y, const T* data, size_t step) const
+        {
+            return saturate_cast<D>(*(const D*)((const char*)data + idx_row_low(y) * step));
+        }
+
+        template <typename T> __device__ __forceinline__ D at_high(int y, const T* data, size_t step) const
+        {
+            return saturate_cast<D>(*(const D*)((const char*)data + idx_row_high(y) * step));
+        }
+
+        template <typename T> __device__ __forceinline__ D at(int y, const T* data, size_t step) const
+        {
+            return saturate_cast<D>(*(const D*)((const char*)data + idx_row(y) * step));
+        }
+
+        int last_row;
+    };
+
+    template <typename D> struct BrdReflect101
+    {
+        typedef D result_type;
+
+        __host__ __device__ __forceinline__ BrdReflect101(int height, int width) : last_row(height - 1), last_col(width - 1) {}
+        template <typename U> __host__ __device__ __forceinline__ BrdReflect101(int height, int width, U) : last_row(height - 1), last_col(width - 1) {}
+
+        __device__ __forceinline__ int idx_row_low(int y) const
+        {
+            return ::abs(y) % (last_row + 1);
+        }
+
+        __device__ __forceinline__ int idx_row_high(int y) const
+        {
+            return ::abs(last_row - ::abs(last_row - y)) % (last_row + 1);
+        }
+
+        __device__ __forceinline__ int idx_row(int y) const
+        {
+            return idx_row_low(idx_row_high(y));
+        }
+
+        __device__ __forceinline__ int idx_col_low(int x) const
+        {
+            return ::abs(x) % (last_col + 1);
+        }
+
+        __device__ __forceinline__ int idx_col_high(int x) const
+        {
+            return ::abs(last_col - ::abs(last_col - x)) % (last_col + 1);
+        }
+
+        __device__ __forceinline__ int idx_col(int x) const
+        {
+            return idx_col_low(idx_col_high(x));
+        }
+
+        template <typename T> __device__ __forceinline__ D at(int y, int x, const T* data, size_t step) const
+        {
+            return saturate_cast<D>(((const T*)((const char*)data + idx_row(y) * step))[idx_col(x)]);
+        }
+
+        template <typename Ptr2D> __device__ __forceinline__ D at(typename Ptr2D::index_type y, typename Ptr2D::index_type x, const Ptr2D& src) const
+        {
+            return saturate_cast<D>(src(idx_row(y), idx_col(x)));
+        }
+
+        int last_row;
+        int last_col;
+    };
+
+    //////////////////////////////////////////////////////////////
+    // BrdReflect
+
+    template <typename D> struct BrdRowReflect
+    {
+        typedef D result_type;
+
+        explicit __host__ __device__ __forceinline__ BrdRowReflect(int width) : last_col(width - 1) {}
+        template <typename U> __host__ __device__ __forceinline__ BrdRowReflect(int width, U) : last_col(width - 1) {}
+
+        __device__ __forceinline__ int idx_col_low(int x) const
+        {
+            return (::abs(x) - (x < 0)) % (last_col + 1);
+        }
+
+        __device__ __forceinline__ int idx_col_high(int x) const
+        {
+            return ::abs(last_col - ::abs(last_col - x) + (x > last_col)) % (last_col + 1);
+        }
+
+        __device__ __forceinline__ int idx_col(int x) const
+        {
+            return idx_col_high(::abs(x) - (x < 0));
+        }
+
+        template <typename T> __device__ __forceinline__ D at_low(int x, const T* data) const
+        {
+            return saturate_cast<D>(data[idx_col_low(x)]);
+        }
+
+        template <typename T> __device__ __forceinline__ D at_high(int x, const T* data) const
+        {
+            return saturate_cast<D>(data[idx_col_high(x)]);
+        }
+
+        template <typename T> __device__ __forceinline__ D at(int x, const T* data) const
+        {
+            return saturate_cast<D>(data[idx_col(x)]);
+        }
+
+        int last_col;
+    };
+
+    template <typename D> struct BrdColReflect
+    {
+        typedef D result_type;
+
+        explicit __host__ __device__ __forceinline__ BrdColReflect(int height) : last_row(height - 1) {}
+        template <typename U> __host__ __device__ __forceinline__ BrdColReflect(int height, U) : last_row(height - 1) {}
+
+        __device__ __forceinline__ int idx_row_low(int y) const
+        {
+            return (::abs(y) - (y < 0)) % (last_row + 1);
+        }
+
+        __device__ __forceinline__ int idx_row_high(int y) const
+        {
+            return ::abs(last_row - ::abs(last_row - y) + (y > last_row)) % (last_row + 1);
+        }
+
+        __device__ __forceinline__ int idx_row(int y) const
+        {
+            return idx_row_high(::abs(y) - (y < 0));
+        }
+
+        template <typename T> __device__ __forceinline__ D at_low(int y, const T* data, size_t step) const
+        {
+            return saturate_cast<D>(*(const D*)((const char*)data + idx_row_low(y) * step));
+        }
+
+        template <typename T> __device__ __forceinline__ D at_high(int y, const T* data, size_t step) const
+        {
+            return saturate_cast<D>(*(const D*)((const char*)data + idx_row_high(y) * step));
+        }
+
+        template <typename T> __device__ __forceinline__ D at(int y, const T* data, size_t step) const
+        {
+            return saturate_cast<D>(*(const D*)((const char*)data + idx_row(y) * step));
+        }
+
+        int last_row;
+    };
+
+    template <typename D> struct BrdReflect
+    {
+        typedef D result_type;
+
+        __host__ __device__ __forceinline__ BrdReflect(int height, int width) : last_row(height - 1), last_col(width - 1) {}
+        template <typename U> __host__ __device__ __forceinline__ BrdReflect(int height, int width, U) : last_row(height - 1), last_col(width - 1) {}
+
+        __device__ __forceinline__ int idx_row_low(int y) const
+        {
+            return (::abs(y) - (y < 0)) % (last_row + 1);
+        }
+
+        __device__ __forceinline__ int idx_row_high(int y) const
+        {
+            return /*::abs*/(last_row - ::abs(last_row - y) + (y > last_row)) /*% (last_row + 1)*/;
+        }
+
+        __device__ __forceinline__ int idx_row(int y) const
+        {
+            return idx_row_low(idx_row_high(y));
+        }
+
+        __device__ __forceinline__ int idx_col_low(int x) const
+        {
+            return (::abs(x) - (x < 0)) % (last_col + 1);
+        }
+
+        __device__ __forceinline__ int idx_col_high(int x) const
+        {
+            return (last_col - ::abs(last_col - x) + (x > last_col));
+        }
+
+        __device__ __forceinline__ int idx_col(int x) const
+        {
+            return idx_col_low(idx_col_high(x));
+        }
+
+        template <typename T> __device__ __forceinline__ D at(int y, int x, const T* data, size_t step) const
+        {
+            return saturate_cast<D>(((const T*)((const char*)data + idx_row(y) * step))[idx_col(x)]);
+        }
+
+        template <typename Ptr2D> __device__ __forceinline__ D at(typename Ptr2D::index_type y, typename Ptr2D::index_type x, const Ptr2D& src) const
+        {
+            return saturate_cast<D>(src(idx_row(y), idx_col(x)));
+        }
+
+        int last_row;
+        int last_col;
+    };
+
+    //////////////////////////////////////////////////////////////
+    // BrdWrap
+
+    template <typename D> struct BrdRowWrap
+    {
+        typedef D result_type;
+
+        explicit __host__ __device__ __forceinline__ BrdRowWrap(int width_) : width(width_) {}
+        template <typename U> __host__ __device__ __forceinline__ BrdRowWrap(int width_, U) : width(width_) {}
+
+        __device__ __forceinline__ int idx_col_low(int x) const
+        {
+            return (x >= 0) * x + (x < 0) * (x - ((x - width + 1) / width) * width);
+        }
+
+        __device__ __forceinline__ int idx_col_high(int x) const
+        {
+            return (x < width) * x + (x >= width) * (x % width);
+        }
+
+        __device__ __forceinline__ int idx_col(int x) const
+        {
+            return idx_col_high(idx_col_low(x));
+        }
+
+        template <typename T> __device__ __forceinline__ D at_low(int x, const T* data) const
+        {
+            return saturate_cast<D>(data[idx_col_low(x)]);
+        }
+
+        template <typename T> __device__ __forceinline__ D at_high(int x, const T* data) const
+        {
+            return saturate_cast<D>(data[idx_col_high(x)]);
+        }
+
+        template <typename T> __device__ __forceinline__ D at(int x, const T* data) const
+        {
+            return saturate_cast<D>(data[idx_col(x)]);
+        }
+
+        int width;
+    };
+
+    template <typename D> struct BrdColWrap
+    {
+        typedef D result_type;
+
+        explicit __host__ __device__ __forceinline__ BrdColWrap(int height_) : height(height_) {}
+        template <typename U> __host__ __device__ __forceinline__ BrdColWrap(int height_, U) : height(height_) {}
+
+        __device__ __forceinline__ int idx_row_low(int y) const
+        {
+            return (y >= 0) * y + (y < 0) * (y - ((y - height + 1) / height) * height);
+        }
+
+        __device__ __forceinline__ int idx_row_high(int y) const
+        {
+            return (y < height) * y + (y >= height) * (y % height);
+        }
+
+        __device__ __forceinline__ int idx_row(int y) const
+        {
+            return idx_row_high(idx_row_low(y));
+        }
+
+        template <typename T> __device__ __forceinline__ D at_low(int y, const T* data, size_t step) const
+        {
+            return saturate_cast<D>(*(const D*)((const char*)data + idx_row_low(y) * step));
+        }
+
+        template <typename T> __device__ __forceinline__ D at_high(int y, const T* data, size_t step) const
+        {
+            return saturate_cast<D>(*(const D*)((const char*)data + idx_row_high(y) * step));
+        }
+
+        template <typename T> __device__ __forceinline__ D at(int y, const T* data, size_t step) const
+        {
+            return saturate_cast<D>(*(const D*)((const char*)data + idx_row(y) * step));
+        }
+
+        int height;
+    };
+
+    template <typename D> struct BrdWrap
+    {
+        typedef D result_type;
+
+        __host__ __device__ __forceinline__ BrdWrap(int height_, int width_) :
+            height(height_), width(width_)
+        {
+        }
+        template <typename U>
+        __host__ __device__ __forceinline__ BrdWrap(int height_, int width_, U) :
+            height(height_), width(width_)
+        {
+        }
+
+        __device__ __forceinline__ int idx_row_low(int y) const
+        {
+            return (y >= 0) ? y : (y - ((y - height + 1) / height) * height);
+        }
+
+        __device__ __forceinline__ int idx_row_high(int y) const
+        {
+            return (y < height) ? y : (y % height);
+        }
+
+        __device__ __forceinline__ int idx_row(int y) const
+        {
+            return idx_row_high(idx_row_low(y));
+        }
+
+        __device__ __forceinline__ int idx_col_low(int x) const
+        {
+            return (x >= 0) ? x : (x - ((x - width + 1) / width) * width);
+        }
+
+        __device__ __forceinline__ int idx_col_high(int x) const
+        {
+            return (x < width) ? x : (x % width);
+        }
+
+        __device__ __forceinline__ int idx_col(int x) const
+        {
+            return idx_col_high(idx_col_low(x));
+        }
+
+        template <typename T> __device__ __forceinline__ D at(int y, int x, const T* data, size_t step) const
+        {
+            return saturate_cast<D>(((const T*)((const char*)data + idx_row(y) * step))[idx_col(x)]);
+        }
+
+        template <typename Ptr2D> __device__ __forceinline__ D at(typename Ptr2D::index_type y, typename Ptr2D::index_type x, const Ptr2D& src) const
+        {
+            return saturate_cast<D>(src(idx_row(y), idx_col(x)));
+        }
+
+        int height;
+        int width;
+    };
+
+    //////////////////////////////////////////////////////////////
+    // BorderReader
+
+    template <typename Ptr2D, typename B> struct BorderReader
+    {
+        typedef typename B::result_type elem_type;
+        typedef typename Ptr2D::index_type index_type;
+
+        __host__ __device__ __forceinline__ BorderReader(const Ptr2D& ptr_, const B& b_) : ptr(ptr_), b(b_) {}
+
+        __device__ __forceinline__ elem_type operator ()(index_type y, index_type x) const
+        {
+            return b.at(y, x, ptr);
+        }
+
+        Ptr2D ptr;
+        B b;
+    };
+
+    // under win32 there is some bug with templated types that passed as kernel parameters
+    // with this specialization all works fine
+    template <typename Ptr2D, typename D> struct BorderReader< Ptr2D, BrdConstant<D> >
+    {
+        typedef typename BrdConstant<D>::result_type elem_type;
+        typedef typename Ptr2D::index_type index_type;
+
+        __host__ __device__ __forceinline__ BorderReader(const Ptr2D& src_, const BrdConstant<D>& b) :
+            src(src_), height(b.height), width(b.width), val(b.val)
+        {
+        }
+
+        __device__ __forceinline__ D operator ()(index_type y, index_type x) const
+        {
+            return (x >= 0 && x < width && y >= 0 && y < height) ? saturate_cast<D>(src(y, x)) : val;
+        }
+
+        Ptr2D src;
+        int height;
+        int width;
+        D val;
+    };
+}}} // namespace cv { namespace cuda { namespace cudev
+
+//! @endcond
+
+#endif // OPENCV_CUDA_BORDER_INTERPOLATE_HPP

model_zoo/official/lite/app/src/main/cpp/include/opencv2/core/cuda/color.hpp

@@ -0,0 +1,309 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CUDA_COLOR_HPP
+#define OPENCV_CUDA_COLOR_HPP
+
+#include "detail/color_detail.hpp"
+
+/** @file
+ * @deprecated Use @ref cudev instead.
+ */
+
+//! @cond IGNORED
+
+namespace cv { namespace cuda { namespace device
+{
+    // All OPENCV_CUDA_IMPLEMENT_*_TRAITS(ColorSpace1_to_ColorSpace2, ...) macros implements
+    // template <typename T> class ColorSpace1_to_ColorSpace2_traits
+    // {
+    //     typedef ... functor_type;
+    //     static __host__ __device__ functor_type create_functor();
+    // };
+
+    OPENCV_CUDA_IMPLEMENT_RGB2RGB_TRAITS(bgr_to_rgb, 3, 3, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2RGB_TRAITS(bgr_to_bgra, 3, 4, 0)
+    OPENCV_CUDA_IMPLEMENT_RGB2RGB_TRAITS(bgr_to_rgba, 3, 4, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2RGB_TRAITS(bgra_to_bgr, 4, 3, 0)
+    OPENCV_CUDA_IMPLEMENT_RGB2RGB_TRAITS(bgra_to_rgb, 4, 3, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2RGB_TRAITS(bgra_to_rgba, 4, 4, 2)
+
+    #undef OPENCV_CUDA_IMPLEMENT_RGB2RGB_TRAITS
+
+    OPENCV_CUDA_IMPLEMENT_RGB2RGB5x5_TRAITS(bgr_to_bgr555, 3, 0, 5)
+    OPENCV_CUDA_IMPLEMENT_RGB2RGB5x5_TRAITS(bgr_to_bgr565, 3, 0, 6)
+    OPENCV_CUDA_IMPLEMENT_RGB2RGB5x5_TRAITS(rgb_to_bgr555, 3, 2, 5)
+    OPENCV_CUDA_IMPLEMENT_RGB2RGB5x5_TRAITS(rgb_to_bgr565, 3, 2, 6)
+    OPENCV_CUDA_IMPLEMENT_RGB2RGB5x5_TRAITS(bgra_to_bgr555, 4, 0, 5)
+    OPENCV_CUDA_IMPLEMENT_RGB2RGB5x5_TRAITS(bgra_to_bgr565, 4, 0, 6)
+    OPENCV_CUDA_IMPLEMENT_RGB2RGB5x5_TRAITS(rgba_to_bgr555, 4, 2, 5)
+    OPENCV_CUDA_IMPLEMENT_RGB2RGB5x5_TRAITS(rgba_to_bgr565, 4, 2, 6)
+
+    #undef OPENCV_CUDA_IMPLEMENT_RGB2RGB5x5_TRAITS
+
+    OPENCV_CUDA_IMPLEMENT_RGB5x52RGB_TRAITS(bgr555_to_rgb, 3, 2, 5)
+    OPENCV_CUDA_IMPLEMENT_RGB5x52RGB_TRAITS(bgr565_to_rgb, 3, 2, 6)
+    OPENCV_CUDA_IMPLEMENT_RGB5x52RGB_TRAITS(bgr555_to_bgr, 3, 0, 5)
+    OPENCV_CUDA_IMPLEMENT_RGB5x52RGB_TRAITS(bgr565_to_bgr, 3, 0, 6)
+    OPENCV_CUDA_IMPLEMENT_RGB5x52RGB_TRAITS(bgr555_to_rgba, 4, 2, 5)
+    OPENCV_CUDA_IMPLEMENT_RGB5x52RGB_TRAITS(bgr565_to_rgba, 4, 2, 6)
+    OPENCV_CUDA_IMPLEMENT_RGB5x52RGB_TRAITS(bgr555_to_bgra, 4, 0, 5)
+    OPENCV_CUDA_IMPLEMENT_RGB5x52RGB_TRAITS(bgr565_to_bgra, 4, 0, 6)
+
+    #undef OPENCV_CUDA_IMPLEMENT_RGB5x52RGB_TRAITS
+
+    OPENCV_CUDA_IMPLEMENT_GRAY2RGB_TRAITS(gray_to_bgr, 3)
+    OPENCV_CUDA_IMPLEMENT_GRAY2RGB_TRAITS(gray_to_bgra, 4)
+
+    #undef OPENCV_CUDA_IMPLEMENT_GRAY2RGB_TRAITS
+
+    OPENCV_CUDA_IMPLEMENT_GRAY2RGB5x5_TRAITS(gray_to_bgr555, 5)
+    OPENCV_CUDA_IMPLEMENT_GRAY2RGB5x5_TRAITS(gray_to_bgr565, 6)
+
+    #undef OPENCV_CUDA_IMPLEMENT_GRAY2RGB5x5_TRAITS
+
+    OPENCV_CUDA_IMPLEMENT_RGB5x52GRAY_TRAITS(bgr555_to_gray, 5)
+    OPENCV_CUDA_IMPLEMENT_RGB5x52GRAY_TRAITS(bgr565_to_gray, 6)
+
+    #undef OPENCV_CUDA_IMPLEMENT_RGB5x52GRAY_TRAITS
+
+    OPENCV_CUDA_IMPLEMENT_RGB2GRAY_TRAITS(rgb_to_gray, 3, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2GRAY_TRAITS(bgr_to_gray, 3, 0)
+    OPENCV_CUDA_IMPLEMENT_RGB2GRAY_TRAITS(rgba_to_gray, 4, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2GRAY_TRAITS(bgra_to_gray, 4, 0)
+
+    #undef OPENCV_CUDA_IMPLEMENT_RGB2GRAY_TRAITS
+
+    OPENCV_CUDA_IMPLEMENT_RGB2YUV_TRAITS(rgb_to_yuv, 3, 3, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2YUV_TRAITS(rgba_to_yuv, 4, 3, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2YUV_TRAITS(rgb_to_yuv4, 3, 4, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2YUV_TRAITS(rgba_to_yuv4, 4, 4, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2YUV_TRAITS(bgr_to_yuv, 3, 3, 0)
+    OPENCV_CUDA_IMPLEMENT_RGB2YUV_TRAITS(bgra_to_yuv, 4, 3, 0)
+    OPENCV_CUDA_IMPLEMENT_RGB2YUV_TRAITS(bgr_to_yuv4, 3, 4, 0)
+    OPENCV_CUDA_IMPLEMENT_RGB2YUV_TRAITS(bgra_to_yuv4, 4, 4, 0)
+
+    #undef OPENCV_CUDA_IMPLEMENT_RGB2YUV_TRAITS
+
+    OPENCV_CUDA_IMPLEMENT_YUV2RGB_TRAITS(yuv_to_rgb, 3, 3, 2)
+    OPENCV_CUDA_IMPLEMENT_YUV2RGB_TRAITS(yuv_to_rgba, 3, 4, 2)
+    OPENCV_CUDA_IMPLEMENT_YUV2RGB_TRAITS(yuv4_to_rgb, 4, 3, 2)
+    OPENCV_CUDA_IMPLEMENT_YUV2RGB_TRAITS(yuv4_to_rgba, 4, 4, 2)
+    OPENCV_CUDA_IMPLEMENT_YUV2RGB_TRAITS(yuv_to_bgr, 3, 3, 0)
+    OPENCV_CUDA_IMPLEMENT_YUV2RGB_TRAITS(yuv_to_bgra, 3, 4, 0)
+    OPENCV_CUDA_IMPLEMENT_YUV2RGB_TRAITS(yuv4_to_bgr, 4, 3, 0)
+    OPENCV_CUDA_IMPLEMENT_YUV2RGB_TRAITS(yuv4_to_bgra, 4, 4, 0)
+
+    #undef OPENCV_CUDA_IMPLEMENT_YUV2RGB_TRAITS
+
+    OPENCV_CUDA_IMPLEMENT_RGB2YCrCb_TRAITS(rgb_to_YCrCb, 3, 3, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2YCrCb_TRAITS(rgba_to_YCrCb, 4, 3, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2YCrCb_TRAITS(rgb_to_YCrCb4, 3, 4, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2YCrCb_TRAITS(rgba_to_YCrCb4, 4, 4, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2YCrCb_TRAITS(bgr_to_YCrCb, 3, 3, 0)
+    OPENCV_CUDA_IMPLEMENT_RGB2YCrCb_TRAITS(bgra_to_YCrCb, 4, 3, 0)
+    OPENCV_CUDA_IMPLEMENT_RGB2YCrCb_TRAITS(bgr_to_YCrCb4, 3, 4, 0)
+    OPENCV_CUDA_IMPLEMENT_RGB2YCrCb_TRAITS(bgra_to_YCrCb4, 4, 4, 0)
+
+    #undef OPENCV_CUDA_IMPLEMENT_RGB2YCrCb_TRAITS
+
+    OPENCV_CUDA_IMPLEMENT_YCrCb2RGB_TRAITS(YCrCb_to_rgb, 3, 3, 2)
+    OPENCV_CUDA_IMPLEMENT_YCrCb2RGB_TRAITS(YCrCb_to_rgba, 3, 4, 2)
+    OPENCV_CUDA_IMPLEMENT_YCrCb2RGB_TRAITS(YCrCb4_to_rgb, 4, 3, 2)
+    OPENCV_CUDA_IMPLEMENT_YCrCb2RGB_TRAITS(YCrCb4_to_rgba, 4, 4, 2)
+    OPENCV_CUDA_IMPLEMENT_YCrCb2RGB_TRAITS(YCrCb_to_bgr, 3, 3, 0)
+    OPENCV_CUDA_IMPLEMENT_YCrCb2RGB_TRAITS(YCrCb_to_bgra, 3, 4, 0)
+    OPENCV_CUDA_IMPLEMENT_YCrCb2RGB_TRAITS(YCrCb4_to_bgr, 4, 3, 0)
+    OPENCV_CUDA_IMPLEMENT_YCrCb2RGB_TRAITS(YCrCb4_to_bgra, 4, 4, 0)
+
+    #undef OPENCV_CUDA_IMPLEMENT_YCrCb2RGB_TRAITS
+
+    OPENCV_CUDA_IMPLEMENT_RGB2XYZ_TRAITS(rgb_to_xyz, 3, 3, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2XYZ_TRAITS(rgba_to_xyz, 4, 3, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2XYZ_TRAITS(rgb_to_xyz4, 3, 4, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2XYZ_TRAITS(rgba_to_xyz4, 4, 4, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2XYZ_TRAITS(bgr_to_xyz, 3, 3, 0)
+    OPENCV_CUDA_IMPLEMENT_RGB2XYZ_TRAITS(bgra_to_xyz, 4, 3, 0)
+    OPENCV_CUDA_IMPLEMENT_RGB2XYZ_TRAITS(bgr_to_xyz4, 3, 4, 0)
+    OPENCV_CUDA_IMPLEMENT_RGB2XYZ_TRAITS(bgra_to_xyz4, 4, 4, 0)
+
+    #undef OPENCV_CUDA_IMPLEMENT_RGB2XYZ_TRAITS
+
+    OPENCV_CUDA_IMPLEMENT_XYZ2RGB_TRAITS(xyz_to_rgb, 3, 3, 2)
+    OPENCV_CUDA_IMPLEMENT_XYZ2RGB_TRAITS(xyz4_to_rgb, 4, 3, 2)
+    OPENCV_CUDA_IMPLEMENT_XYZ2RGB_TRAITS(xyz_to_rgba, 3, 4, 2)
+    OPENCV_CUDA_IMPLEMENT_XYZ2RGB_TRAITS(xyz4_to_rgba, 4, 4, 2)
+    OPENCV_CUDA_IMPLEMENT_XYZ2RGB_TRAITS(xyz_to_bgr, 3, 3, 0)
+    OPENCV_CUDA_IMPLEMENT_XYZ2RGB_TRAITS(xyz4_to_bgr, 4, 3, 0)
+    OPENCV_CUDA_IMPLEMENT_XYZ2RGB_TRAITS(xyz_to_bgra, 3, 4, 0)
+    OPENCV_CUDA_IMPLEMENT_XYZ2RGB_TRAITS(xyz4_to_bgra, 4, 4, 0)
+
+    #undef OPENCV_CUDA_IMPLEMENT_XYZ2RGB_TRAITS
+
+    OPENCV_CUDA_IMPLEMENT_RGB2HSV_TRAITS(rgb_to_hsv, 3, 3, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2HSV_TRAITS(rgba_to_hsv, 4, 3, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2HSV_TRAITS(rgb_to_hsv4, 3, 4, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2HSV_TRAITS(rgba_to_hsv4, 4, 4, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2HSV_TRAITS(bgr_to_hsv, 3, 3, 0)
+    OPENCV_CUDA_IMPLEMENT_RGB2HSV_TRAITS(bgra_to_hsv, 4, 3, 0)
+    OPENCV_CUDA_IMPLEMENT_RGB2HSV_TRAITS(bgr_to_hsv4, 3, 4, 0)
+    OPENCV_CUDA_IMPLEMENT_RGB2HSV_TRAITS(bgra_to_hsv4, 4, 4, 0)
+
+    #undef OPENCV_CUDA_IMPLEMENT_RGB2HSV_TRAITS
+
+    OPENCV_CUDA_IMPLEMENT_HSV2RGB_TRAITS(hsv_to_rgb, 3, 3, 2)
+    OPENCV_CUDA_IMPLEMENT_HSV2RGB_TRAITS(hsv_to_rgba, 3, 4, 2)
+    OPENCV_CUDA_IMPLEMENT_HSV2RGB_TRAITS(hsv4_to_rgb, 4, 3, 2)
+    OPENCV_CUDA_IMPLEMENT_HSV2RGB_TRAITS(hsv4_to_rgba, 4, 4, 2)
+    OPENCV_CUDA_IMPLEMENT_HSV2RGB_TRAITS(hsv_to_bgr, 3, 3, 0)
+    OPENCV_CUDA_IMPLEMENT_HSV2RGB_TRAITS(hsv_to_bgra, 3, 4, 0)
+    OPENCV_CUDA_IMPLEMENT_HSV2RGB_TRAITS(hsv4_to_bgr, 4, 3, 0)
+    OPENCV_CUDA_IMPLEMENT_HSV2RGB_TRAITS(hsv4_to_bgra, 4, 4, 0)
+
+    #undef OPENCV_CUDA_IMPLEMENT_HSV2RGB_TRAITS
+
+    OPENCV_CUDA_IMPLEMENT_RGB2HLS_TRAITS(rgb_to_hls, 3, 3, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2HLS_TRAITS(rgba_to_hls, 4, 3, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2HLS_TRAITS(rgb_to_hls4, 3, 4, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2HLS_TRAITS(rgba_to_hls4, 4, 4, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2HLS_TRAITS(bgr_to_hls, 3, 3, 0)
+    OPENCV_CUDA_IMPLEMENT_RGB2HLS_TRAITS(bgra_to_hls, 4, 3, 0)
+    OPENCV_CUDA_IMPLEMENT_RGB2HLS_TRAITS(bgr_to_hls4, 3, 4, 0)
+    OPENCV_CUDA_IMPLEMENT_RGB2HLS_TRAITS(bgra_to_hls4, 4, 4, 0)
+
+    #undef OPENCV_CUDA_IMPLEMENT_RGB2HLS_TRAITS
+
+    OPENCV_CUDA_IMPLEMENT_HLS2RGB_TRAITS(hls_to_rgb, 3, 3, 2)
+    OPENCV_CUDA_IMPLEMENT_HLS2RGB_TRAITS(hls_to_rgba, 3, 4, 2)
+    OPENCV_CUDA_IMPLEMENT_HLS2RGB_TRAITS(hls4_to_rgb, 4, 3, 2)
+    OPENCV_CUDA_IMPLEMENT_HLS2RGB_TRAITS(hls4_to_rgba, 4, 4, 2)
+    OPENCV_CUDA_IMPLEMENT_HLS2RGB_TRAITS(hls_to_bgr, 3, 3, 0)
+    OPENCV_CUDA_IMPLEMENT_HLS2RGB_TRAITS(hls_to_bgra, 3, 4, 0)
+    OPENCV_CUDA_IMPLEMENT_HLS2RGB_TRAITS(hls4_to_bgr, 4, 3, 0)
+    OPENCV_CUDA_IMPLEMENT_HLS2RGB_TRAITS(hls4_to_bgra, 4, 4, 0)
+
+    #undef OPENCV_CUDA_IMPLEMENT_HLS2RGB_TRAITS
+
+    OPENCV_CUDA_IMPLEMENT_RGB2Lab_TRAITS(rgb_to_lab, 3, 3, true, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2Lab_TRAITS(rgba_to_lab, 4, 3, true, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2Lab_TRAITS(rgb_to_lab4, 3, 4, true, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2Lab_TRAITS(rgba_to_lab4, 4, 4, true, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2Lab_TRAITS(bgr_to_lab, 3, 3, true, 0)
+    OPENCV_CUDA_IMPLEMENT_RGB2Lab_TRAITS(bgra_to_lab, 4, 3, true, 0)
+    OPENCV_CUDA_IMPLEMENT_RGB2Lab_TRAITS(bgr_to_lab4, 3, 4, true, 0)
+    OPENCV_CUDA_IMPLEMENT_RGB2Lab_TRAITS(bgra_to_lab4, 4, 4, true, 0)
+
+    OPENCV_CUDA_IMPLEMENT_RGB2Lab_TRAITS(lrgb_to_lab, 3, 3, false, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2Lab_TRAITS(lrgba_to_lab, 4, 3, false, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2Lab_TRAITS(lrgb_to_lab4, 3, 4, false, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2Lab_TRAITS(lrgba_to_lab4, 4, 4, false, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2Lab_TRAITS(lbgr_to_lab, 3, 3, false, 0)
+    OPENCV_CUDA_IMPLEMENT_RGB2Lab_TRAITS(lbgra_to_lab, 4, 3, false, 0)
+    OPENCV_CUDA_IMPLEMENT_RGB2Lab_TRAITS(lbgr_to_lab4, 3, 4, false, 0)
+    OPENCV_CUDA_IMPLEMENT_RGB2Lab_TRAITS(lbgra_to_lab4, 4, 4, false, 0)
+
+    #undef OPENCV_CUDA_IMPLEMENT_RGB2Lab_TRAITS
+
+    OPENCV_CUDA_IMPLEMENT_Lab2RGB_TRAITS(lab_to_rgb, 3, 3, true, 2)
+    OPENCV_CUDA_IMPLEMENT_Lab2RGB_TRAITS(lab4_to_rgb, 4, 3, true, 2)
+    OPENCV_CUDA_IMPLEMENT_Lab2RGB_TRAITS(lab_to_rgba, 3, 4, true, 2)
+    OPENCV_CUDA_IMPLEMENT_Lab2RGB_TRAITS(lab4_to_rgba, 4, 4, true, 2)
+    OPENCV_CUDA_IMPLEMENT_Lab2RGB_TRAITS(lab_to_bgr, 3, 3, true, 0)
+    OPENCV_CUDA_IMPLEMENT_Lab2RGB_TRAITS(lab4_to_bgr, 4, 3, true, 0)
+    OPENCV_CUDA_IMPLEMENT_Lab2RGB_TRAITS(lab_to_bgra, 3, 4, true, 0)
+    OPENCV_CUDA_IMPLEMENT_Lab2RGB_TRAITS(lab4_to_bgra, 4, 4, true, 0)
+
+    OPENCV_CUDA_IMPLEMENT_Lab2RGB_TRAITS(lab_to_lrgb, 3, 3, false, 2)
+    OPENCV_CUDA_IMPLEMENT_Lab2RGB_TRAITS(lab4_to_lrgb, 4, 3, false, 2)
+    OPENCV_CUDA_IMPLEMENT_Lab2RGB_TRAITS(lab_to_lrgba, 3, 4, false, 2)
+    OPENCV_CUDA_IMPLEMENT_Lab2RGB_TRAITS(lab4_to_lrgba, 4, 4, false, 2)
+    OPENCV_CUDA_IMPLEMENT_Lab2RGB_TRAITS(lab_to_lbgr, 3, 3, false, 0)
+    OPENCV_CUDA_IMPLEMENT_Lab2RGB_TRAITS(lab4_to_lbgr, 4, 3, false, 0)
+    OPENCV_CUDA_IMPLEMENT_Lab2RGB_TRAITS(lab_to_lbgra, 3, 4, false, 0)
+    OPENCV_CUDA_IMPLEMENT_Lab2RGB_TRAITS(lab4_to_lbgra, 4, 4, false, 0)
+
+    #undef OPENCV_CUDA_IMPLEMENT_Lab2RGB_TRAITS
+
+    OPENCV_CUDA_IMPLEMENT_RGB2Luv_TRAITS(rgb_to_luv, 3, 3, true, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2Luv_TRAITS(rgba_to_luv, 4, 3, true, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2Luv_TRAITS(rgb_to_luv4, 3, 4, true, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2Luv_TRAITS(rgba_to_luv4, 4, 4, true, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2Luv_TRAITS(bgr_to_luv, 3, 3, true, 0)
+    OPENCV_CUDA_IMPLEMENT_RGB2Luv_TRAITS(bgra_to_luv, 4, 3, true, 0)
+    OPENCV_CUDA_IMPLEMENT_RGB2Luv_TRAITS(bgr_to_luv4, 3, 4, true, 0)
+    OPENCV_CUDA_IMPLEMENT_RGB2Luv_TRAITS(bgra_to_luv4, 4, 4, true, 0)
+
+    OPENCV_CUDA_IMPLEMENT_RGB2Luv_TRAITS(lrgb_to_luv, 3, 3, false, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2Luv_TRAITS(lrgba_to_luv, 4, 3, false, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2Luv_TRAITS(lrgb_to_luv4, 3, 4, false, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2Luv_TRAITS(lrgba_to_luv4, 4, 4, false, 2)
+    OPENCV_CUDA_IMPLEMENT_RGB2Luv_TRAITS(lbgr_to_luv, 3, 3, false, 0)
+    OPENCV_CUDA_IMPLEMENT_RGB2Luv_TRAITS(lbgra_to_luv, 4, 3, false, 0)
+    OPENCV_CUDA_IMPLEMENT_RGB2Luv_TRAITS(lbgr_to_luv4, 3, 4, false, 0)
+    OPENCV_CUDA_IMPLEMENT_RGB2Luv_TRAITS(lbgra_to_luv4, 4, 4, false, 0)
+
+    #undef OPENCV_CUDA_IMPLEMENT_RGB2Luv_TRAITS
+
+    OPENCV_CUDA_IMPLEMENT_Luv2RGB_TRAITS(luv_to_rgb, 3, 3, true, 2)
+    OPENCV_CUDA_IMPLEMENT_Luv2RGB_TRAITS(luv4_to_rgb, 4, 3, true, 2)
+    OPENCV_CUDA_IMPLEMENT_Luv2RGB_TRAITS(luv_to_rgba, 3, 4, true, 2)
+    OPENCV_CUDA_IMPLEMENT_Luv2RGB_TRAITS(luv4_to_rgba, 4, 4, true, 2)
+    OPENCV_CUDA_IMPLEMENT_Luv2RGB_TRAITS(luv_to_bgr, 3, 3, true, 0)
+    OPENCV_CUDA_IMPLEMENT_Luv2RGB_TRAITS(luv4_to_bgr, 4, 3, true, 0)
+    OPENCV_CUDA_IMPLEMENT_Luv2RGB_TRAITS(luv_to_bgra, 3, 4, true, 0)
+    OPENCV_CUDA_IMPLEMENT_Luv2RGB_TRAITS(luv4_to_bgra, 4, 4, true, 0)
+
+    OPENCV_CUDA_IMPLEMENT_Luv2RGB_TRAITS(luv_to_lrgb, 3, 3, false, 2)
+    OPENCV_CUDA_IMPLEMENT_Luv2RGB_TRAITS(luv4_to_lrgb, 4, 3, false, 2)
+    OPENCV_CUDA_IMPLEMENT_Luv2RGB_TRAITS(luv_to_lrgba, 3, 4, false, 2)
+    OPENCV_CUDA_IMPLEMENT_Luv2RGB_TRAITS(luv4_to_lrgba, 4, 4, false, 2)
+    OPENCV_CUDA_IMPLEMENT_Luv2RGB_TRAITS(luv_to_lbgr, 3, 3, false, 0)
+    OPENCV_CUDA_IMPLEMENT_Luv2RGB_TRAITS(luv4_to_lbgr, 4, 3, false, 0)
+    OPENCV_CUDA_IMPLEMENT_Luv2RGB_TRAITS(luv_to_lbgra, 3, 4, false, 0)
+    OPENCV_CUDA_IMPLEMENT_Luv2RGB_TRAITS(luv4_to_lbgra, 4, 4, false, 0)
+
+    #undef OPENCV_CUDA_IMPLEMENT_Luv2RGB_TRAITS
+}}} // namespace cv { namespace cuda { namespace cudev
+
+//! @endcond
+
+#endif // OPENCV_CUDA_COLOR_HPP

model_zoo/official/lite/app/src/main/cpp/include/opencv2/core/cuda/common.hpp

@@ -0,0 +1,123 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CUDA_COMMON_HPP
+#define OPENCV_CUDA_COMMON_HPP
+
+#include <cuda_runtime.h>
+#include "opencv2/core/cuda_types.hpp"
+#include "opencv2/core/cvdef.h"
+#include "opencv2/core/base.hpp"
+
+/** @file
+ * @deprecated Use @ref cudev instead.
+ */
+
+//! @cond IGNORED
+
+#ifndef CV_PI_F
+    #ifndef CV_PI
+        #define CV_PI_F 3.14159265f
+    #else
+        #define CV_PI_F ((float)CV_PI)
+    #endif
+#endif
+
+namespace cv { namespace cuda {
+    static inline void checkCudaError(cudaError_t err, const char* file, const int line, const char* func)
+    {
+        if (cudaSuccess != err)
+            cv::error(cv::Error::GpuApiCallError, cudaGetErrorString(err), func, file, line);
+    }
+}}
+
+#ifndef cudaSafeCall
+    #define cudaSafeCall(expr)  cv::cuda::checkCudaError(expr, __FILE__, __LINE__, CV_Func)
+#endif
+
+namespace cv { namespace cuda
+{
+    template <typename T> static inline bool isAligned(const T* ptr, size_t size)
+    {
+        return reinterpret_cast<size_t>(ptr) % size == 0;
+    }
+
+    static inline bool isAligned(size_t step, size_t size)
+    {
+        return step % size == 0;
+    }
+}}
+
+namespace cv { namespace cuda
+{
+    namespace device
+    {
+        __host__ __device__ __forceinline__ int divUp(int total, int grain)
+        {
+            return (total + grain - 1) / grain;
+        }
+
+        template<class T> inline void bindTexture(const textureReference* tex, const PtrStepSz<T>& img)
+        {
+            cudaChannelFormatDesc desc = cudaCreateChannelDesc<T>();
+            cudaSafeCall( cudaBindTexture2D(0, tex, img.ptr(), &desc, img.cols, img.rows, img.step) );
+        }
+
+        template<class T> inline void createTextureObjectPitch2D(cudaTextureObject_t* tex, PtrStepSz<T>& img, const cudaTextureDesc& texDesc)
+        {
+            cudaResourceDesc resDesc;
+            memset(&resDesc, 0, sizeof(resDesc));
+            resDesc.resType = cudaResourceTypePitch2D;
+            resDesc.res.pitch2D.devPtr = static_cast<void*>(img.ptr());
+            resDesc.res.pitch2D.height = img.rows;
+            resDesc.res.pitch2D.width = img.cols;
+            resDesc.res.pitch2D.pitchInBytes = img.step;
+            resDesc.res.pitch2D.desc = cudaCreateChannelDesc<T>();
+
+            cudaSafeCall( cudaCreateTextureObject(tex, &resDesc, &texDesc, NULL) );
+        }
+    }
+}}
+
+//! @endcond
+
+#endif // OPENCV_CUDA_COMMON_HPP

model_zoo/official/lite/app/src/main/cpp/include/opencv2/core/cuda/datamov_utils.hpp

@@ -0,0 +1,113 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CUDA_DATAMOV_UTILS_HPP
+#define OPENCV_CUDA_DATAMOV_UTILS_HPP
+
+#include "common.hpp"
+
+/** @file
+ * @deprecated Use @ref cudev instead.
+ */
+
+//! @cond IGNORED
+
+namespace cv { namespace cuda { namespace device
+{
+    #if defined __CUDA_ARCH__ && __CUDA_ARCH__ >= 200
+
+        // for Fermi memory space is detected automatically
+        template <typename T> struct ForceGlob
+        {
+            __device__ __forceinline__ static void Load(const T* ptr, int offset, T& val)  { val = ptr[offset];  }
+        };
+
+    #else // __CUDA_ARCH__ >= 200
+
+        #if defined(_WIN64) || defined(__LP64__)
+            // 64-bit register modifier for inlined asm
+            #define OPENCV_CUDA_ASM_PTR "l"
+        #else
+            // 32-bit register modifier for inlined asm
+            #define OPENCV_CUDA_ASM_PTR "r"
+        #endif
+
+        template<class T> struct ForceGlob;
+
+        #define OPENCV_CUDA_DEFINE_FORCE_GLOB(base_type, ptx_type, reg_mod) \
+            template <> struct ForceGlob<base_type> \
+            { \
+                __device__ __forceinline__ static void Load(const base_type* ptr, int offset, base_type& val) \
+                { \
+                    asm("ld.global."#ptx_type" %0, [%1];" : "="#reg_mod(val) : OPENCV_CUDA_ASM_PTR(ptr + offset)); \
+                } \
+            };
+
+        #define OPENCV_CUDA_DEFINE_FORCE_GLOB_B(base_type, ptx_type) \
+            template <> struct ForceGlob<base_type> \
+            { \
+                __device__ __forceinline__ static void Load(const base_type* ptr, int offset, base_type& val) \
+                { \
+                    asm("ld.global."#ptx_type" %0, [%1];" : "=r"(*reinterpret_cast<uint*>(&val)) : OPENCV_CUDA_ASM_PTR(ptr + offset)); \
+                } \
+            };
+
+            OPENCV_CUDA_DEFINE_FORCE_GLOB_B(uchar,  u8)
+            OPENCV_CUDA_DEFINE_FORCE_GLOB_B(schar,  s8)
+            OPENCV_CUDA_DEFINE_FORCE_GLOB_B(char,   b8)
+            OPENCV_CUDA_DEFINE_FORCE_GLOB  (ushort, u16, h)
+            OPENCV_CUDA_DEFINE_FORCE_GLOB  (short,  s16, h)
+            OPENCV_CUDA_DEFINE_FORCE_GLOB  (uint,   u32, r)
+            OPENCV_CUDA_DEFINE_FORCE_GLOB  (int,    s32, r)
+            OPENCV_CUDA_DEFINE_FORCE_GLOB  (float,  f32, f)
+            OPENCV_CUDA_DEFINE_FORCE_GLOB  (double, f64, d)
+
+        #undef OPENCV_CUDA_DEFINE_FORCE_GLOB
+        #undef OPENCV_CUDA_DEFINE_FORCE_GLOB_B
+        #undef OPENCV_CUDA_ASM_PTR
+
+    #endif // __CUDA_ARCH__ >= 200
+}}} // namespace cv { namespace cuda { namespace cudev
+
+//! @endcond
+
+#endif // OPENCV_CUDA_DATAMOV_UTILS_HPP

model_zoo/official/lite/app/src/main/cpp/include/opencv2/core/cuda/detail/reduce.hpp

@@ -0,0 +1,365 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CUDA_REDUCE_DETAIL_HPP
+#define OPENCV_CUDA_REDUCE_DETAIL_HPP
+
+#include <thrust/tuple.h>
+#include "../warp.hpp"
+#include "../warp_shuffle.hpp"
+
+//! @cond IGNORED
+
+namespace cv { namespace cuda { namespace device
+{
+    namespace reduce_detail
+    {
+        template <typename T> struct GetType;
+        template <typename T> struct GetType<T*>
+        {
+            typedef T type;
+        };
+        template <typename T> struct GetType<volatile T*>
+        {
+            typedef T type;
+        };
+        template <typename T> struct GetType<T&>
+        {
+            typedef T type;
+        };
+
+        template <unsigned int I, unsigned int N>
+        struct For
+        {
+            template <class PointerTuple, class ValTuple>
+            static __device__ void loadToSmem(const PointerTuple& smem, const ValTuple& val, unsigned int tid)
+            {
+                thrust::get<I>(smem)[tid] = thrust::get<I>(val);
+
+                For<I + 1, N>::loadToSmem(smem, val, tid);
+            }
+            template <class PointerTuple, class ValTuple>
+            static __device__ void loadFromSmem(const PointerTuple& smem, const ValTuple& val, unsigned int tid)
+            {
+                thrust::get<I>(val) = thrust::get<I>(smem)[tid];
+
+                For<I + 1, N>::loadFromSmem(smem, val, tid);
+            }
+
+            template <class PointerTuple, class ValTuple, class OpTuple>
+            static __device__ void merge(const PointerTuple& smem, const ValTuple& val, unsigned int tid, unsigned int delta, const OpTuple& op)
+            {
+                typename GetType<typename thrust::tuple_element<I, PointerTuple>::type>::type reg = thrust::get<I>(smem)[tid + delta];
+                thrust::get<I>(smem)[tid] = thrust::get<I>(val) = thrust::get<I>(op)(thrust::get<I>(val), reg);
+
+                For<I + 1, N>::merge(smem, val, tid, delta, op);
+            }
+            template <class ValTuple, class OpTuple>
+            static __device__ void mergeShfl(const ValTuple& val, unsigned int delta, unsigned int width, const OpTuple& op)
+            {
+                typename GetType<typename thrust::tuple_element<I, ValTuple>::type>::type reg = shfl_down(thrust::get<I>(val), delta, width);
+                thrust::get<I>(val) = thrust::get<I>(op)(thrust::get<I>(val), reg);
+
+                For<I + 1, N>::mergeShfl(val, delta, width, op);
+            }
+        };
+        template <unsigned int N>
+        struct For<N, N>
+        {
+            template <class PointerTuple, class ValTuple>
+            static __device__ void loadToSmem(const PointerTuple&, const ValTuple&, unsigned int)
+            {
+            }
+            template <class PointerTuple, class ValTuple>
+            static __device__ void loadFromSmem(const PointerTuple&, const ValTuple&, unsigned int)
+            {
+            }
+
+            template <class PointerTuple, class ValTuple, class OpTuple>
+            static __device__ void merge(const PointerTuple&, const ValTuple&, unsigned int, unsigned int, const OpTuple&)
+            {
+            }
+            template <class ValTuple, class OpTuple>
+            static __device__ void mergeShfl(const ValTuple&, unsigned int, unsigned int, const OpTuple&)
+            {
+            }
+        };
+
+        template <typename T>
+        __device__ __forceinline__ void loadToSmem(volatile T* smem, T& val, unsigned int tid)
+        {
+            smem[tid] = val;
+        }
+        template <typename T>
+        __device__ __forceinline__ void loadFromSmem(volatile T* smem, T& val, unsigned int tid)
+        {
+            val = smem[tid];
+        }
+        template <typename P0, typename P1, typename P2, typename P3, typename P4, typename P5, typename P6, typename P7, typename P8, typename P9,
+                  typename R0, typename R1, typename R2, typename R3, typename R4, typename R5, typename R6, typename R7, typename R8, typename R9>
+        __device__ __forceinline__ void loadToSmem(const thrust::tuple<P0, P1, P2, P3, P4, P5, P6, P7, P8, P9>& smem,
+                                                       const thrust::tuple<R0, R1, R2, R3, R4, R5, R6, R7, R8, R9>& val,
+                                                       unsigned int tid)
+        {
+            For<0, thrust::tuple_size<thrust::tuple<P0, P1, P2, P3, P4, P5, P6, P7, P8, P9> >::value>::loadToSmem(smem, val, tid);
+        }
+        template <typename P0, typename P1, typename P2, typename P3, typename P4, typename P5, typename P6, typename P7, typename P8, typename P9,
+                  typename R0, typename R1, typename R2, typename R3, typename R4, typename R5, typename R6, typename R7, typename R8, typename R9>
+        __device__ __forceinline__ void loadFromSmem(const thrust::tuple<P0, P1, P2, P3, P4, P5, P6, P7, P8, P9>& smem,
+                                                         const thrust::tuple<R0, R1, R2, R3, R4, R5, R6, R7, R8, R9>& val,
+                                                         unsigned int tid)
+        {
+            For<0, thrust::tuple_size<thrust::tuple<P0, P1, P2, P3, P4, P5, P6, P7, P8, P9> >::value>::loadFromSmem(smem, val, tid);
+        }
+
+        template <typename T, class Op>
+        __device__ __forceinline__ void merge(volatile T* smem, T& val, unsigned int tid, unsigned int delta, const Op& op)
+        {
+            T reg = smem[tid + delta];
+            smem[tid] = val = op(val, reg);
+        }
+        template <typename T, class Op>
+        __device__ __forceinline__ void mergeShfl(T& val, unsigned int delta, unsigned int width, const Op& op)
+        {
+            T reg = shfl_down(val, delta, width);
+            val = op(val, reg);
+        }
+        template <typename P0, typename P1, typename P2, typename P3, typename P4, typename P5, typename P6, typename P7, typename P8, typename P9,
+                  typename R0, typename R1, typename R2, typename R3, typename R4, typename R5, typename R6, typename R7, typename R8, typename R9,
+                  class Op0, class Op1, class Op2, class Op3, class Op4, class Op5, class Op6, class Op7, class Op8, class Op9>
+        __device__ __forceinline__ void merge(const thrust::tuple<P0, P1, P2, P3, P4, P5, P6, P7, P8, P9>& smem,
+                                              const thrust::tuple<R0, R1, R2, R3, R4, R5, R6, R7, R8, R9>& val,
+                                              unsigned int tid,
+                                              unsigned int delta,
+                                              const thrust::tuple<Op0, Op1, Op2, Op3, Op4, Op5, Op6, Op7, Op8, Op9>& op)
+        {
+            For<0, thrust::tuple_size<thrust::tuple<P0, P1, P2, P3, P4, P5, P6, P7, P8, P9> >::value>::merge(smem, val, tid, delta, op);
+        }
+        template <typename R0, typename R1, typename R2, typename R3, typename R4, typename R5, typename R6, typename R7, typename R8, typename R9,
+                  class Op0, class Op1, class Op2, class Op3, class Op4, class Op5, class Op6, class Op7, class Op8, class Op9>
+        __device__ __forceinline__ void mergeShfl(const thrust::tuple<R0, R1, R2, R3, R4, R5, R6, R7, R8, R9>& val,
+                                                  unsigned int delta,
+                                                  unsigned int width,
+                                                  const thrust::tuple<Op0, Op1, Op2, Op3, Op4, Op5, Op6, Op7, Op8, Op9>& op)
+        {
+            For<0, thrust::tuple_size<thrust::tuple<R0, R1, R2, R3, R4, R5, R6, R7, R8, R9> >::value>::mergeShfl(val, delta, width, op);
+        }
+
+        template <unsigned int N> struct Generic
+        {
+            template <typename Pointer, typename Reference, class Op>
+            static __device__ void reduce(Pointer smem, Reference val, unsigned int tid, Op op)
+            {
+                loadToSmem(smem, val, tid);
+                if (N >= 32)
+                    __syncthreads();
+
+                if (N >= 2048)
+                {
+                    if (tid < 1024)
+                        merge(smem, val, tid, 1024, op);
+
+                    __syncthreads();
+                }
+                if (N >= 1024)
+                {
+                    if (tid < 512)
+                        merge(smem, val, tid, 512, op);
+
+                    __syncthreads();
+                }
+                if (N >= 512)
+                {
+                    if (tid < 256)
+                        merge(smem, val, tid, 256, op);
+
+                    __syncthreads();
+                }
+                if (N >= 256)
+                {
+                    if (tid < 128)
+                        merge(smem, val, tid, 128, op);
+
+                    __syncthreads();
+                }
+                if (N >= 128)
+                {
+                    if (tid < 64)
+                        merge(smem, val, tid, 64, op);
+
+                    __syncthreads();
+                }
+                if (N >= 64)
+                {
+                    if (tid < 32)
+                        merge(smem, val, tid, 32, op);
+                }
+
+                if (tid < 16)
+                {
+                    merge(smem, val, tid, 16, op);
+                    merge(smem, val, tid, 8, op);
+                    merge(smem, val, tid, 4, op);
+                    merge(smem, val, tid, 2, op);
+                    merge(smem, val, tid, 1, op);
+                }
+            }
+        };
+
+        template <unsigned int I, typename Pointer, typename Reference, class Op>
+        struct Unroll
+        {
+            static __device__ void loopShfl(Reference val, Op op, unsigned int N)
+            {
+                mergeShfl(val, I, N, op);
+                Unroll<I / 2, Pointer, Reference, Op>::loopShfl(val, op, N);
+            }
+            static __device__ void loop(Pointer smem, Reference val, unsigned int tid, Op op)
+            {
+                merge(smem, val, tid, I, op);
+                Unroll<I / 2, Pointer, Reference, Op>::loop(smem, val, tid, op);
+            }
+        };
+        template <typename Pointer, typename Reference, class Op>
+        struct Unroll<0, Pointer, Reference, Op>
+        {
+            static __device__ void loopShfl(Reference, Op, unsigned int)
+            {
+            }
+            static __device__ void loop(Pointer, Reference, unsigned int, Op)
+            {
+            }
+        };
+
+        template <unsigned int N> struct WarpOptimized
+        {
+            template <typename Pointer, typename Reference, class Op>
+            static __device__ void reduce(Pointer smem, Reference val, unsigned int tid, Op op)
+            {
+            #if defined __CUDA_ARCH__ && __CUDA_ARCH__ >= 300
+                CV_UNUSED(smem);
+                CV_UNUSED(tid);
+
+                Unroll<N / 2, Pointer, Reference, Op>::loopShfl(val, op, N);
+            #else
+                loadToSmem(smem, val, tid);
+
+                if (tid < N / 2)
+                    Unroll<N / 2, Pointer, Reference, Op>::loop(smem, val, tid, op);
+            #endif
+            }
+        };
+
+        template <unsigned int N> struct GenericOptimized32
+        {
+            enum { M = N / 32 };
+
+            template <typename Pointer, typename Reference, class Op>
+            static __device__ void reduce(Pointer smem, Reference val, unsigned int tid, Op op)
+            {
+                const unsigned int laneId = Warp::laneId();
+
+            #if defined __CUDA_ARCH__ && __CUDA_ARCH__ >= 300
+                Unroll<16, Pointer, Reference, Op>::loopShfl(val, op, warpSize);
+
+                if (laneId == 0)
+                    loadToSmem(smem, val, tid / 32);
+            #else
+                loadToSmem(smem, val, tid);
+
+                if (laneId < 16)
+                    Unroll<16, Pointer, Reference, Op>::loop(smem, val, tid, op);
+
+                __syncthreads();
+
+                if (laneId == 0)
+                    loadToSmem(smem, val, tid / 32);
+            #endif
+
+                __syncthreads();
+
+                loadFromSmem(smem, val, tid);
+
+                if (tid < 32)
+                {
+                #if defined __CUDA_ARCH__ && __CUDA_ARCH__ >= 300
+                    Unroll<M / 2, Pointer, Reference, Op>::loopShfl(val, op, M);
+                #else
+                    Unroll<M / 2, Pointer, Reference, Op>::loop(smem, val, tid, op);
+                #endif
+                }
+            }
+        };
+
+        template <bool val, class T1, class T2> struct StaticIf;
+        template <class T1, class T2> struct StaticIf<true, T1, T2>
+        {
+            typedef T1 type;
+        };
+        template <class T1, class T2> struct StaticIf<false, T1, T2>
+        {
+            typedef T2 type;
+        };
+
+        template <unsigned int N> struct IsPowerOf2
+        {
+            enum { value = ((N != 0) && !(N & (N - 1))) };
+        };
+
+        template <unsigned int N> struct Dispatcher
+        {
+            typedef typename StaticIf<
+                (N <= 32) && IsPowerOf2<N>::value,
+                WarpOptimized<N>,
+                typename StaticIf<
+                    (N <= 1024) && IsPowerOf2<N>::value,
+                    GenericOptimized32<N>,
+                    Generic<N>
+                >::type
+            >::type reductor;
+        };
+    }
+}}}
+
+//! @endcond
+
+#endif // OPENCV_CUDA_REDUCE_DETAIL_HPP

model_zoo/official/lite/app/src/main/cpp/include/opencv2/core/cuda/detail/reduce_key_val.hpp

@@ -0,0 +1,502 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CUDA_PRED_VAL_REDUCE_DETAIL_HPP
+#define OPENCV_CUDA_PRED_VAL_REDUCE_DETAIL_HPP
+
+#include <thrust/tuple.h>
+#include "../warp.hpp"
+#include "../warp_shuffle.hpp"
+
+//! @cond IGNORED
+
+namespace cv { namespace cuda { namespace device
+{
+    namespace reduce_key_val_detail
+    {
+        template <typename T> struct GetType;
+        template <typename T> struct GetType<T*>
+        {
+            typedef T type;
+        };
+        template <typename T> struct GetType<volatile T*>
+        {
+            typedef T type;
+        };
+        template <typename T> struct GetType<T&>
+        {
+            typedef T type;
+        };
+
+        template <unsigned int I, unsigned int N>
+        struct For
+        {
+            template <class PointerTuple, class ReferenceTuple>
+            static __device__ void loadToSmem(const PointerTuple& smem, const ReferenceTuple& data, unsigned int tid)
+            {
+                thrust::get<I>(smem)[tid] = thrust::get<I>(data);
+
+                For<I + 1, N>::loadToSmem(smem, data, tid);
+            }
+            template <class PointerTuple, class ReferenceTuple>
+            static __device__ void loadFromSmem(const PointerTuple& smem, const ReferenceTuple& data, unsigned int tid)
+            {
+                thrust::get<I>(data) = thrust::get<I>(smem)[tid];
+
+                For<I + 1, N>::loadFromSmem(smem, data, tid);
+            }
+
+            template <class ReferenceTuple>
+            static __device__ void copyShfl(const ReferenceTuple& val, unsigned int delta, int width)
+            {
+                thrust::get<I>(val) = shfl_down(thrust::get<I>(val), delta, width);
+
+                For<I + 1, N>::copyShfl(val, delta, width);
+            }
+            template <class PointerTuple, class ReferenceTuple>
+            static __device__ void copy(const PointerTuple& svals, const ReferenceTuple& val, unsigned int tid, unsigned int delta)
+            {
+                thrust::get<I>(svals)[tid] = thrust::get<I>(val) = thrust::get<I>(svals)[tid + delta];
+
+                For<I + 1, N>::copy(svals, val, tid, delta);
+            }
+
+            template <class KeyReferenceTuple, class ValReferenceTuple, class CmpTuple>
+            static __device__ void mergeShfl(const KeyReferenceTuple& key, const ValReferenceTuple& val, const CmpTuple& cmp, unsigned int delta, int width)
+            {
+                typename GetType<typename thrust::tuple_element<I, KeyReferenceTuple>::type>::type reg = shfl_down(thrust::get<I>(key), delta, width);
+
+                if (thrust::get<I>(cmp)(reg, thrust::get<I>(key)))
+                {
+                    thrust::get<I>(key) = reg;
+                    thrust::get<I>(val) = shfl_down(thrust::get<I>(val), delta, width);
+                }
+
+                For<I + 1, N>::mergeShfl(key, val, cmp, delta, width);
+            }
+            template <class KeyPointerTuple, class KeyReferenceTuple, class ValPointerTuple, class ValReferenceTuple, class CmpTuple>
+            static __device__ void merge(const KeyPointerTuple& skeys, const KeyReferenceTuple& key,
+                                         const ValPointerTuple& svals, const ValReferenceTuple& val,
+                                         const CmpTuple& cmp,
+                                         unsigned int tid, unsigned int delta)
+            {
+                typename GetType<typename thrust::tuple_element<I, KeyPointerTuple>::type>::type reg = thrust::get<I>(skeys)[tid + delta];
+
+                if (thrust::get<I>(cmp)(reg, thrust::get<I>(key)))
+                {
+                    thrust::get<I>(skeys)[tid] = thrust::get<I>(key) = reg;
+                    thrust::get<I>(svals)[tid] = thrust::get<I>(val) = thrust::get<I>(svals)[tid + delta];
+                }
+
+                For<I + 1, N>::merge(skeys, key, svals, val, cmp, tid, delta);
+            }
+        };
+        template <unsigned int N>
+        struct For<N, N>
+        {
+            template <class PointerTuple, class ReferenceTuple>
+            static __device__ void loadToSmem(const PointerTuple&, const ReferenceTuple&, unsigned int)
+            {
+            }
+            template <class PointerTuple, class ReferenceTuple>
+            static __device__ void loadFromSmem(const PointerTuple&, const ReferenceTuple&, unsigned int)
+            {
+            }
+
+            template <class ReferenceTuple>
+            static __device__ void copyShfl(const ReferenceTuple&, unsigned int, int)
+            {
+            }
+            template <class PointerTuple, class ReferenceTuple>
+            static __device__ void copy(const PointerTuple&, const ReferenceTuple&, unsigned int, unsigned int)
+            {
+            }
+
+            template <class KeyReferenceTuple, class ValReferenceTuple, class CmpTuple>
+            static __device__ void mergeShfl(const KeyReferenceTuple&, const ValReferenceTuple&, const CmpTuple&, unsigned int, int)
+            {
+            }
+            template <class KeyPointerTuple, class KeyReferenceTuple, class ValPointerTuple, class ValReferenceTuple, class CmpTuple>
+            static __device__ void merge(const KeyPointerTuple&, const KeyReferenceTuple&,
+                                         const ValPointerTuple&, const ValReferenceTuple&,
+                                         const CmpTuple&,
+                                         unsigned int, unsigned int)
+            {
+            }
+        };
+
+        //////////////////////////////////////////////////////
+        // loadToSmem
+
+        template <typename T>
+        __device__ __forceinline__ void loadToSmem(volatile T* smem, T& data, unsigned int tid)
+        {
+            smem[tid] = data;
+        }
+        template <typename T>
+        __device__ __forceinline__ void loadFromSmem(volatile T* smem, T& data, unsigned int tid)
+        {
+            data = smem[tid];
+        }
+        template <typename VP0, typename VP1, typename VP2, typename VP3, typename VP4, typename VP5, typename VP6, typename VP7, typename VP8, typename VP9,
+                  typename VR0, typename VR1, typename VR2, typename VR3, typename VR4, typename VR5, typename VR6, typename VR7, typename VR8, typename VR9>
+        __device__ __forceinline__ void loadToSmem(const thrust::tuple<VP0, VP1, VP2, VP3, VP4, VP5, VP6, VP7, VP8, VP9>& smem,
+                                                   const thrust::tuple<VR0, VR1, VR2, VR3, VR4, VR5, VR6, VR7, VR8, VR9>& data,
+                                                   unsigned int tid)
+        {
+            For<0, thrust::tuple_size<thrust::tuple<VP0, VP1, VP2, VP3, VP4, VP5, VP6, VP7, VP8, VP9> >::value>::loadToSmem(smem, data, tid);
+        }
+        template <typename VP0, typename VP1, typename VP2, typename VP3, typename VP4, typename VP5, typename VP6, typename VP7, typename VP8, typename VP9,
+                  typename VR0, typename VR1, typename VR2, typename VR3, typename VR4, typename VR5, typename VR6, typename VR7, typename VR8, typename VR9>
+        __device__ __forceinline__ void loadFromSmem(const thrust::tuple<VP0, VP1, VP2, VP3, VP4, VP5, VP6, VP7, VP8, VP9>& smem,
+                                                     const thrust::tuple<VR0, VR1, VR2, VR3, VR4, VR5, VR6, VR7, VR8, VR9>& data,
+                                                     unsigned int tid)
+        {
+            For<0, thrust::tuple_size<thrust::tuple<VP0, VP1, VP2, VP3, VP4, VP5, VP6, VP7, VP8, VP9> >::value>::loadFromSmem(smem, data, tid);
+        }
+
+        //////////////////////////////////////////////////////
+        // copyVals
+
+        template <typename V>
+        __device__ __forceinline__ void copyValsShfl(V& val, unsigned int delta, int width)
+        {
+            val = shfl_down(val, delta, width);
+        }
+        template <typename V>
+        __device__ __forceinline__ void copyVals(volatile V* svals, V& val, unsigned int tid, unsigned int delta)
+        {
+            svals[tid] = val = svals[tid + delta];
+        }
+        template <typename VR0, typename VR1, typename VR2, typename VR3, typename VR4, typename VR5, typename VR6, typename VR7, typename VR8, typename VR9>
+        __device__ __forceinline__ void copyValsShfl(const thrust::tuple<VR0, VR1, VR2, VR3, VR4, VR5, VR6, VR7, VR8, VR9>& val,
+                                                     unsigned int delta,
+                                                     int width)
+        {
+            For<0, thrust::tuple_size<thrust::tuple<VR0, VR1, VR2, VR3, VR4, VR5, VR6, VR7, VR8, VR9> >::value>::copyShfl(val, delta, width);
+        }
+        template <typename VP0, typename VP1, typename VP2, typename VP3, typename VP4, typename VP5, typename VP6, typename VP7, typename VP8, typename VP9,
+                  typename VR0, typename VR1, typename VR2, typename VR3, typename VR4, typename VR5, typename VR6, typename VR7, typename VR8, typename VR9>
+        __device__ __forceinline__ void copyVals(const thrust::tuple<VP0, VP1, VP2, VP3, VP4, VP5, VP6, VP7, VP8, VP9>& svals,
+                                                 const thrust::tuple<VR0, VR1, VR2, VR3, VR4, VR5, VR6, VR7, VR8, VR9>& val,
+                                                 unsigned int tid, unsigned int delta)
+        {
+            For<0, thrust::tuple_size<thrust::tuple<VP0, VP1, VP2, VP3, VP4, VP5, VP6, VP7, VP8, VP9> >::value>::copy(svals, val, tid, delta);
+        }
+
+        //////////////////////////////////////////////////////
+        // merge
+
+        template <typename K, typename V, class Cmp>
+        __device__ __forceinline__ void mergeShfl(K& key, V& val, const Cmp& cmp, unsigned int delta, int width)
+        {
+            K reg = shfl_down(key, delta, width);
+
+            if (cmp(reg, key))
+            {
+                key = reg;
+                copyValsShfl(val, delta, width);
+            }
+        }
+        template <typename K, typename V, class Cmp>
+        __device__ __forceinline__ void merge(volatile K* skeys, K& key, volatile V* svals, V& val, const Cmp& cmp, unsigned int tid, unsigned int delta)
+        {
+            K reg = skeys[tid + delta];
+
+            if (cmp(reg, key))
+            {
+                skeys[tid] = key = reg;
+                copyVals(svals, val, tid, delta);
+            }
+        }
+        template <typename K,
+                  typename VR0, typename VR1, typename VR2, typename VR3, typename VR4, typename VR5, typename VR6, typename VR7, typename VR8, typename VR9,
+                  class Cmp>
+        __device__ __forceinline__ void mergeShfl(K& key,
+                                                  const thrust::tuple<VR0, VR1, VR2, VR3, VR4, VR5, VR6, VR7, VR8, VR9>& val,
+                                                  const Cmp& cmp,
+                                                  unsigned int delta, int width)
+        {
+            K reg = shfl_down(key, delta, width);
+
+            if (cmp(reg, key))
+            {
+                key = reg;
+                copyValsShfl(val, delta, width);
+            }
+        }
+        template <typename K,
+                  typename VP0, typename VP1, typename VP2, typename VP3, typename VP4, typename VP5, typename VP6, typename VP7, typename VP8, typename VP9,
+                  typename VR0, typename VR1, typename VR2, typename VR3, typename VR4, typename VR5, typename VR6, typename VR7, typename VR8, typename VR9,
+                  class Cmp>
+        __device__ __forceinline__ void merge(volatile K* skeys, K& key,
+                                              const thrust::tuple<VP0, VP1, VP2, VP3, VP4, VP5, VP6, VP7, VP8, VP9>& svals,
+                                              const thrust::tuple<VR0, VR1, VR2, VR3, VR4, VR5, VR6, VR7, VR8, VR9>& val,
+                                              const Cmp& cmp, unsigned int tid, unsigned int delta)
+        {
+            K reg = skeys[tid + delta];
+
+            if (cmp(reg, key))
+            {
+                skeys[tid] = key = reg;
+                copyVals(svals, val, tid, delta);
+            }
+        }
+        template <typename KR0, typename KR1, typename KR2, typename KR3, typename KR4, typename KR5, typename KR6, typename KR7, typename KR8, typename KR9,
+                  typename VR0, typename VR1, typename VR2, typename VR3, typename VR4, typename VR5, typename VR6, typename VR7, typename VR8, typename VR9,
+                  class Cmp0, class Cmp1, class Cmp2, class Cmp3, class Cmp4, class Cmp5, class Cmp6, class Cmp7, class Cmp8, class Cmp9>
+        __device__ __forceinline__ void mergeShfl(const thrust::tuple<KR0, KR1, KR2, KR3, KR4, KR5, KR6, KR7, KR8, KR9>& key,
+                                                  const thrust::tuple<VR0, VR1, VR2, VR3, VR4, VR5, VR6, VR7, VR8, VR9>& val,
+                                                  const thrust::tuple<Cmp0, Cmp1, Cmp2, Cmp3, Cmp4, Cmp5, Cmp6, Cmp7, Cmp8, Cmp9>& cmp,
+                                                  unsigned int delta, int width)
+        {
+            For<0, thrust::tuple_size<thrust::tuple<KR0, KR1, KR2, KR3, KR4, KR5, KR6, KR7, KR8, KR9> >::value>::mergeShfl(key, val, cmp, delta, width);
+        }
+        template <typename KP0, typename KP1, typename KP2, typename KP3, typename KP4, typename KP5, typename KP6, typename KP7, typename KP8, typename KP9,
+                  typename KR0, typename KR1, typename KR2, typename KR3, typename KR4, typename KR5, typename KR6, typename KR7, typename KR8, typename KR9,
+                  typename VP0, typename VP1, typename VP2, typename VP3, typename VP4, typename VP5, typename VP6, typename VP7, typename VP8, typename VP9,
+                  typename VR0, typename VR1, typename VR2, typename VR3, typename VR4, typename VR5, typename VR6, typename VR7, typename VR8, typename VR9,
+                  class Cmp0, class Cmp1, class Cmp2, class Cmp3, class Cmp4, class Cmp5, class Cmp6, class Cmp7, class Cmp8, class Cmp9>
+        __device__ __forceinline__ void merge(const thrust::tuple<KP0, KP1, KP2, KP3, KP4, KP5, KP6, KP7, KP8, KP9>& skeys,
+                                              const thrust::tuple<KR0, KR1, KR2, KR3, KR4, KR5, KR6, KR7, KR8, KR9>& key,
+                                              const thrust::tuple<VP0, VP1, VP2, VP3, VP4, VP5, VP6, VP7, VP8, VP9>& svals,
+                                              const thrust::tuple<VR0, VR1, VR2, VR3, VR4, VR5, VR6, VR7, VR8, VR9>& val,
+                                              const thrust::tuple<Cmp0, Cmp1, Cmp2, Cmp3, Cmp4, Cmp5, Cmp6, Cmp7, Cmp8, Cmp9>& cmp,
+                                              unsigned int tid, unsigned int delta)
+        {
+            For<0, thrust::tuple_size<thrust::tuple<VP0, VP1, VP2, VP3, VP4, VP5, VP6, VP7, VP8, VP9> >::value>::merge(skeys, key, svals, val, cmp, tid, delta);
+        }
+
+        //////////////////////////////////////////////////////
+        // Generic
+
+        template <unsigned int N> struct Generic
+        {
+            template <class KP, class KR, class VP, class VR, class Cmp>
+            static __device__ void reduce(KP skeys, KR key, VP svals, VR val, unsigned int tid, Cmp cmp)
+            {
+                loadToSmem(skeys, key, tid);
+                loadValsToSmem(svals, val, tid);
+                if (N >= 32)
+                    __syncthreads();
+
+                if (N >= 2048)
+                {
+                    if (tid < 1024)
+                        merge(skeys, key, svals, val, cmp, tid, 1024);
+
+                    __syncthreads();
+                }
+                if (N >= 1024)
+                {
+                    if (tid < 512)
+                        merge(skeys, key, svals, val, cmp, tid, 512);
+
+                    __syncthreads();
+                }
+                if (N >= 512)
+                {
+                    if (tid < 256)
+                        merge(skeys, key, svals, val, cmp, tid, 256);
+
+                    __syncthreads();
+                }
+                if (N >= 256)
+                {
+                    if (tid < 128)
+                        merge(skeys, key, svals, val, cmp, tid, 128);
+
+                    __syncthreads();
+                }
+                if (N >= 128)
+                {
+                    if (tid < 64)
+                        merge(skeys, key, svals, val, cmp, tid, 64);
+
+                    __syncthreads();
+                }
+                if (N >= 64)
+                {
+                    if (tid < 32)
+                        merge(skeys, key, svals, val, cmp, tid, 32);
+                }
+
+                if (tid < 16)
+                {
+                    merge(skeys, key, svals, val, cmp, tid, 16);
+                    merge(skeys, key, svals, val, cmp, tid, 8);
+                    merge(skeys, key, svals, val, cmp, tid, 4);
+                    merge(skeys, key, svals, val, cmp, tid, 2);
+                    merge(skeys, key, svals, val, cmp, tid, 1);
+                }
+            }
+        };
+
+        template <unsigned int I, class KP, class KR, class VP, class VR, class Cmp>
+        struct Unroll
+        {
+            static __device__ void loopShfl(KR key, VR val, Cmp cmp, unsigned int N)
+            {
+                mergeShfl(key, val, cmp, I, N);
+                Unroll<I / 2, KP, KR, VP, VR, Cmp>::loopShfl(key, val, cmp, N);
+            }
+            static __device__ void loop(KP skeys, KR key, VP svals, VR val, unsigned int tid, Cmp cmp)
+            {
+                merge(skeys, key, svals, val, cmp, tid, I);
+                Unroll<I / 2, KP, KR, VP, VR, Cmp>::loop(skeys, key, svals, val, tid, cmp);
+            }
+        };
+        template <class KP, class KR, class VP, class VR, class Cmp>
+        struct Unroll<0, KP, KR, VP, VR, Cmp>
+        {
+            static __device__ void loopShfl(KR, VR, Cmp, unsigned int)
+            {
+            }
+            static __device__ void loop(KP, KR, VP, VR, unsigned int, Cmp)
+            {
+            }
+        };
+
+        template <unsigned int N> struct WarpOptimized
+        {
+            template <class KP, class KR, class VP, class VR, class Cmp>
+            static __device__ void reduce(KP skeys, KR key, VP svals, VR val, unsigned int tid, Cmp cmp)
+            {
+            #if 0 // __CUDA_ARCH__ >= 300
+                CV_UNUSED(skeys);
+                CV_UNUSED(svals);
+                CV_UNUSED(tid);
+
+                Unroll<N / 2, KP, KR, VP, VR, Cmp>::loopShfl(key, val, cmp, N);
+            #else
+                loadToSmem(skeys, key, tid);
+                loadToSmem(svals, val, tid);
+
+                if (tid < N / 2)
+                    Unroll<N / 2, KP, KR, VP, VR, Cmp>::loop(skeys, key, svals, val, tid, cmp);
+            #endif
+            }
+        };
+
+        template <unsigned int N> struct GenericOptimized32
+        {
+            enum { M = N / 32 };
+
+            template <class KP, class KR, class VP, class VR, class Cmp>
+            static __device__ void reduce(KP skeys, KR key, VP svals, VR val, unsigned int tid, Cmp cmp)
+            {
+                const unsigned int laneId = Warp::laneId();
+
+            #if 0 // __CUDA_ARCH__ >= 300
+                Unroll<16, KP, KR, VP, VR, Cmp>::loopShfl(key, val, cmp, warpSize);
+
+                if (laneId == 0)
+                {
+                    loadToSmem(skeys, key, tid / 32);
+                    loadToSmem(svals, val, tid / 32);
+                }
+            #else
+                loadToSmem(skeys, key, tid);
+                loadToSmem(svals, val, tid);
+
+                if (laneId < 16)
+                    Unroll<16, KP, KR, VP, VR, Cmp>::loop(skeys, key, svals, val, tid, cmp);
+
+                __syncthreads();
+
+                if (laneId == 0)
+                {
+                    loadToSmem(skeys, key, tid / 32);
+                    loadToSmem(svals, val, tid / 32);
+                }
+            #endif
+
+                __syncthreads();
+
+                loadFromSmem(skeys, key, tid);
+
+                if (tid < 32)
+                {
+                #if 0 // __CUDA_ARCH__ >= 300
+                    loadFromSmem(svals, val, tid);
+
+                    Unroll<M / 2, KP, KR, VP, VR, Cmp>::loopShfl(key, val, cmp, M);
+                #else
+                    Unroll<M / 2, KP, KR, VP, VR, Cmp>::loop(skeys, key, svals, val, tid, cmp);
+                #endif
+                }
+            }
+        };
+
+        template <bool val, class T1, class T2> struct StaticIf;
+        template <class T1, class T2> struct StaticIf<true, T1, T2>
+        {
+            typedef T1 type;
+        };
+        template <class T1, class T2> struct StaticIf<false, T1, T2>
+        {
+            typedef T2 type;
+        };
+
+        template <unsigned int N> struct IsPowerOf2
+        {
+            enum { value = ((N != 0) && !(N & (N - 1))) };
+        };
+
+        template <unsigned int N> struct Dispatcher
+        {
+            typedef typename StaticIf<
+                (N <= 32) && IsPowerOf2<N>::value,
+                WarpOptimized<N>,
+                typename StaticIf<
+                    (N <= 1024) && IsPowerOf2<N>::value,
+                    GenericOptimized32<N>,
+                    Generic<N>
+                >::type
+            >::type reductor;
+        };
+    }
+}}}
+
+//! @endcond
+
+#endif // OPENCV_CUDA_PRED_VAL_REDUCE_DETAIL_HPP

model_zoo/official/lite/app/src/main/cpp/include/opencv2/core/cuda/detail/transform_detail.hpp

@@ -0,0 +1,392 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CUDA_TRANSFORM_DETAIL_HPP
+#define OPENCV_CUDA_TRANSFORM_DETAIL_HPP
+
+#include "../common.hpp"
+#include "../vec_traits.hpp"
+#include "../functional.hpp"
+
+//! @cond IGNORED
+
+namespace cv { namespace cuda { namespace device
+{
+    namespace transform_detail
+    {
+        //! Read Write Traits
+
+        template <typename T, typename D, int shift> struct UnaryReadWriteTraits
+        {
+            typedef typename TypeVec<T, shift>::vec_type read_type;
+            typedef typename TypeVec<D, shift>::vec_type write_type;
+        };
+
+        template <typename T1, typename T2, typename D, int shift> struct BinaryReadWriteTraits
+        {
+            typedef typename TypeVec<T1, shift>::vec_type read_type1;
+            typedef typename TypeVec<T2, shift>::vec_type read_type2;
+            typedef typename TypeVec<D, shift>::vec_type write_type;
+        };
+
+        //! Transform kernels
+
+        template <int shift> struct OpUnroller;
+        template <> struct OpUnroller<1>
+        {
+            template <typename T, typename D, typename UnOp, typename Mask>
+            static __device__ __forceinline__ void unroll(const T& src, D& dst, const Mask& mask, UnOp& op, int x_shifted, int y)
+            {
+                if (mask(y, x_shifted))
+                    dst.x = op(src.x);
+            }
+
+            template <typename T1, typename T2, typename D, typename BinOp, typename Mask>
+            static __device__ __forceinline__ void unroll(const T1& src1, const T2& src2, D& dst, const Mask& mask, BinOp& op, int x_shifted, int y)
+            {
+                if (mask(y, x_shifted))
+                    dst.x = op(src1.x, src2.x);
+            }
+        };
+        template <> struct OpUnroller<2>
+        {
+            template <typename T, typename D, typename UnOp, typename Mask>
+            static __device__ __forceinline__ void unroll(const T& src, D& dst, const Mask& mask, UnOp& op, int x_shifted, int y)
+            {
+                if (mask(y, x_shifted))
+                    dst.x = op(src.x);
+                if (mask(y, x_shifted + 1))
+                    dst.y = op(src.y);
+            }
+
+            template <typename T1, typename T2, typename D, typename BinOp, typename Mask>
+            static __device__ __forceinline__ void unroll(const T1& src1, const T2& src2, D& dst, const Mask& mask, BinOp& op, int x_shifted, int y)
+            {
+                if (mask(y, x_shifted))
+                    dst.x = op(src1.x, src2.x);
+                if (mask(y, x_shifted + 1))
+                    dst.y = op(src1.y, src2.y);
+            }
+        };
+        template <> struct OpUnroller<3>
+        {
+            template <typename T, typename D, typename UnOp, typename Mask>
+            static __device__ __forceinline__ void unroll(const T& src, D& dst, const Mask& mask, const UnOp& op, int x_shifted, int y)
+            {
+                if (mask(y, x_shifted))
+                    dst.x = op(src.x);
+                if (mask(y, x_shifted + 1))
+                    dst.y = op(src.y);
+                if (mask(y, x_shifted + 2))
+                    dst.z = op(src.z);
+            }
+
+            template <typename T1, typename T2, typename D, typename BinOp, typename Mask>
+            static __device__ __forceinline__ void unroll(const T1& src1, const T2& src2, D& dst, const Mask& mask, const BinOp& op, int x_shifted, int y)
+            {
+                if (mask(y, x_shifted))
+                    dst.x = op(src1.x, src2.x);
+                if (mask(y, x_shifted + 1))
+                    dst.y = op(src1.y, src2.y);
+                if (mask(y, x_shifted + 2))
+                    dst.z = op(src1.z, src2.z);
+            }
+        };
+        template <> struct OpUnroller<4>
+        {
+            template <typename T, typename D, typename UnOp, typename Mask>
+            static __device__ __forceinline__ void unroll(const T& src, D& dst, const Mask& mask, const UnOp& op, int x_shifted, int y)
+            {
+                if (mask(y, x_shifted))
+                    dst.x = op(src.x);
+                if (mask(y, x_shifted + 1))
+                    dst.y = op(src.y);
+                if (mask(y, x_shifted + 2))
+                    dst.z = op(src.z);
+                if (mask(y, x_shifted + 3))
+                    dst.w = op(src.w);
+            }
+
+            template <typename T1, typename T2, typename D, typename BinOp, typename Mask>
+            static __device__ __forceinline__ void unroll(const T1& src1, const T2& src2, D& dst, const Mask& mask, const BinOp& op, int x_shifted, int y)
+            {
+                if (mask(y, x_shifted))
+                    dst.x = op(src1.x, src2.x);
+                if (mask(y, x_shifted + 1))
+                    dst.y = op(src1.y, src2.y);
+                if (mask(y, x_shifted + 2))
+                    dst.z = op(src1.z, src2.z);
+                if (mask(y, x_shifted + 3))
+                    dst.w = op(src1.w, src2.w);
+            }
+        };
+        template <> struct OpUnroller<8>
+        {
+            template <typename T, typename D, typename UnOp, typename Mask>
+            static __device__ __forceinline__ void unroll(const T& src, D& dst, const Mask& mask, const UnOp& op, int x_shifted, int y)
+            {
+                if (mask(y, x_shifted))
+                    dst.a0 = op(src.a0);
+                if (mask(y, x_shifted + 1))
+                    dst.a1 = op(src.a1);
+                if (mask(y, x_shifted + 2))
+                    dst.a2 = op(src.a2);
+                if (mask(y, x_shifted + 3))
+                    dst.a3 = op(src.a3);
+                if (mask(y, x_shifted + 4))
+                    dst.a4 = op(src.a4);
+                if (mask(y, x_shifted + 5))
+                    dst.a5 = op(src.a5);
+                if (mask(y, x_shifted + 6))
+                    dst.a6 = op(src.a6);
+                if (mask(y, x_shifted + 7))
+                    dst.a7 = op(src.a7);
+            }
+
+            template <typename T1, typename T2, typename D, typename BinOp, typename Mask>
+            static __device__ __forceinline__ void unroll(const T1& src1, const T2& src2, D& dst, const Mask& mask, const BinOp& op, int x_shifted, int y)
+            {
+                if (mask(y, x_shifted))
+                    dst.a0 = op(src1.a0, src2.a0);
+                if (mask(y, x_shifted + 1))
+                    dst.a1 = op(src1.a1, src2.a1);
+                if (mask(y, x_shifted + 2))
+                    dst.a2 = op(src1.a2, src2.a2);
+                if (mask(y, x_shifted + 3))
+                    dst.a3 = op(src1.a3, src2.a3);
+                if (mask(y, x_shifted + 4))
+                    dst.a4 = op(src1.a4, src2.a4);
+                if (mask(y, x_shifted + 5))
+                    dst.a5 = op(src1.a5, src2.a5);
+                if (mask(y, x_shifted + 6))
+                    dst.a6 = op(src1.a6, src2.a6);
+                if (mask(y, x_shifted + 7))
+                    dst.a7 = op(src1.a7, src2.a7);
+            }
+        };
+
+        template <typename T, typename D, typename UnOp, typename Mask>
+        static __global__ void transformSmart(const PtrStepSz<T> src_, PtrStep<D> dst_, const Mask mask, const UnOp op)
+        {
+            typedef TransformFunctorTraits<UnOp> ft;
+            typedef typename UnaryReadWriteTraits<T, D, ft::smart_shift>::read_type read_type;
+            typedef typename UnaryReadWriteTraits<T, D, ft::smart_shift>::write_type write_type;
+
+            const int x = threadIdx.x + blockIdx.x * blockDim.x;
+            const int y = threadIdx.y + blockIdx.y * blockDim.y;
+            const int x_shifted = x * ft::smart_shift;
+
+            if (y < src_.rows)
+            {
+                const T* src = src_.ptr(y);
+                D* dst = dst_.ptr(y);
+
+                if (x_shifted + ft::smart_shift - 1 < src_.cols)
+                {
+                    const read_type src_n_el = ((const read_type*)src)[x];
+                    OpUnroller<ft::smart_shift>::unroll(src_n_el, ((write_type*)dst)[x], mask, op, x_shifted, y);
+                }
+                else
+                {
+                    for (int real_x = x_shifted; real_x < src_.cols; ++real_x)
+                    {
+                        if (mask(y, real_x))
+                            dst[real_x] = op(src[real_x]);
+                    }
+                }
+            }
+        }
+
+        template <typename T, typename D, typename UnOp, typename Mask>
+        __global__ static void transformSimple(const PtrStepSz<T> src, PtrStep<D> dst, const Mask mask, const UnOp op)
+        {
+            const int x = blockDim.x * blockIdx.x + threadIdx.x;
+            const int y = blockDim.y * blockIdx.y + threadIdx.y;
+
+            if (x < src.cols && y < src.rows && mask(y, x))
+            {
+                dst.ptr(y)[x] = op(src.ptr(y)[x]);
+            }
+        }
+
+        template <typename T1, typename T2, typename D, typename BinOp, typename Mask>
+        static __global__ void transformSmart(const PtrStepSz<T1> src1_, const PtrStep<T2> src2_, PtrStep<D> dst_,
+            const Mask mask, const BinOp op)
+        {
+            typedef TransformFunctorTraits<BinOp> ft;
+            typedef typename BinaryReadWriteTraits<T1, T2, D, ft::smart_shift>::read_type1 read_type1;
+            typedef typename BinaryReadWriteTraits<T1, T2, D, ft::smart_shift>::read_type2 read_type2;
+            typedef typename BinaryReadWriteTraits<T1, T2, D, ft::smart_shift>::write_type write_type;
+
+            const int x = threadIdx.x + blockIdx.x * blockDim.x;
+            const int y = threadIdx.y + blockIdx.y * blockDim.y;
+            const int x_shifted = x * ft::smart_shift;
+
+            if (y < src1_.rows)
+            {
+                const T1* src1 = src1_.ptr(y);
+                const T2* src2 = src2_.ptr(y);
+                D* dst = dst_.ptr(y);
+
+                if (x_shifted + ft::smart_shift - 1 < src1_.cols)
+                {
+                    const read_type1 src1_n_el = ((const read_type1*)src1)[x];
+                    const read_type2 src2_n_el = ((const read_type2*)src2)[x];
+
+                    OpUnroller<ft::smart_shift>::unroll(src1_n_el, src2_n_el, ((write_type*)dst)[x], mask, op, x_shifted, y);
+                }
+                else
+                {
+                    for (int real_x = x_shifted; real_x < src1_.cols; ++real_x)
+                    {
+                        if (mask(y, real_x))
+                            dst[real_x] = op(src1[real_x], src2[real_x]);
+                    }
+                }
+            }
+        }
+
+        template <typename T1, typename T2, typename D, typename BinOp, typename Mask>
+        static __global__ void transformSimple(const PtrStepSz<T1> src1, const PtrStep<T2> src2, PtrStep<D> dst,
+            const Mask mask, const BinOp op)
+        {
+            const int x = blockDim.x * blockIdx.x + threadIdx.x;
+            const int y = blockDim.y * blockIdx.y + threadIdx.y;
+
+            if (x < src1.cols && y < src1.rows && mask(y, x))
+            {
+                const T1 src1_data = src1.ptr(y)[x];
+                const T2 src2_data = src2.ptr(y)[x];
+                dst.ptr(y)[x] = op(src1_data, src2_data);
+            }
+        }
+
+        template <bool UseSmart> struct TransformDispatcher;
+        template<> struct TransformDispatcher<false>
+        {
+            template <typename T, typename D, typename UnOp, typename Mask>
+            static void call(PtrStepSz<T> src, PtrStepSz<D> dst, UnOp op, Mask mask, cudaStream_t stream)
+            {
+                typedef TransformFunctorTraits<UnOp> ft;
+
+                const dim3 threads(ft::simple_block_dim_x, ft::simple_block_dim_y, 1);
+                const dim3 grid(divUp(src.cols, threads.x), divUp(src.rows, threads.y), 1);
+
+                transformSimple<T, D><<<grid, threads, 0, stream>>>(src, dst, mask, op);
+                cudaSafeCall( cudaGetLastError() );
+
+                if (stream == 0)
+                    cudaSafeCall( cudaDeviceSynchronize() );
+            }
+
+            template <typename T1, typename T2, typename D, typename BinOp, typename Mask>
+            static void call(PtrStepSz<T1> src1, PtrStepSz<T2> src2, PtrStepSz<D> dst, BinOp op, Mask mask, cudaStream_t stream)
+            {
+                typedef TransformFunctorTraits<BinOp> ft;
+
+                const dim3 threads(ft::simple_block_dim_x, ft::simple_block_dim_y, 1);
+                const dim3 grid(divUp(src1.cols, threads.x), divUp(src1.rows, threads.y), 1);
+
+                transformSimple<T1, T2, D><<<grid, threads, 0, stream>>>(src1, src2, dst, mask, op);
+                cudaSafeCall( cudaGetLastError() );
+
+                if (stream == 0)
+                    cudaSafeCall( cudaDeviceSynchronize() );
+            }
+        };
+        template<> struct TransformDispatcher<true>
+        {
+            template <typename T, typename D, typename UnOp, typename Mask>
+            static void call(PtrStepSz<T> src, PtrStepSz<D> dst, UnOp op, Mask mask, cudaStream_t stream)
+            {
+                typedef TransformFunctorTraits<UnOp> ft;
+
+                CV_StaticAssert(ft::smart_shift != 1, "");
+
+                if (!isAligned(src.data, ft::smart_shift * sizeof(T)) || !isAligned(src.step, ft::smart_shift * sizeof(T)) ||
+                    !isAligned(dst.data, ft::smart_shift * sizeof(D)) || !isAligned(dst.step, ft::smart_shift * sizeof(D)))
+                {
+                    TransformDispatcher<false>::call(src, dst, op, mask, stream);
+                    return;
+                }
+
+                const dim3 threads(ft::smart_block_dim_x, ft::smart_block_dim_y, 1);
+                const dim3 grid(divUp(src.cols, threads.x * ft::smart_shift), divUp(src.rows, threads.y), 1);
+
+                transformSmart<T, D><<<grid, threads, 0, stream>>>(src, dst, mask, op);
+                cudaSafeCall( cudaGetLastError() );
+
+                if (stream == 0)
+                    cudaSafeCall( cudaDeviceSynchronize() );
+            }
+
+            template <typename T1, typename T2, typename D, typename BinOp, typename Mask>
+            static void call(PtrStepSz<T1> src1, PtrStepSz<T2> src2, PtrStepSz<D> dst, BinOp op, Mask mask, cudaStream_t stream)
+            {
+                typedef TransformFunctorTraits<BinOp> ft;
+
+                CV_StaticAssert(ft::smart_shift != 1, "");
+
+                if (!isAligned(src1.data, ft::smart_shift * sizeof(T1)) || !isAligned(src1.step, ft::smart_shift * sizeof(T1)) ||
+                    !isAligned(src2.data, ft::smart_shift * sizeof(T2)) || !isAligned(src2.step, ft::smart_shift * sizeof(T2)) ||
+                    !isAligned(dst.data, ft::smart_shift * sizeof(D)) || !isAligned(dst.step, ft::smart_shift * sizeof(D)))
+                {
+                    TransformDispatcher<false>::call(src1, src2, dst, op, mask, stream);
+                    return;
+                }
+
+                const dim3 threads(ft::smart_block_dim_x, ft::smart_block_dim_y, 1);
+                const dim3 grid(divUp(src1.cols, threads.x * ft::smart_shift), divUp(src1.rows, threads.y), 1);
+
+                transformSmart<T1, T2, D><<<grid, threads, 0, stream>>>(src1, src2, dst, mask, op);
+                cudaSafeCall( cudaGetLastError() );
+
+                if (stream == 0)
+                    cudaSafeCall( cudaDeviceSynchronize() );
+            }
+        };
+    } // namespace transform_detail
+}}} // namespace cv { namespace cuda { namespace cudev
+
+//! @endcond
+
+#endif // OPENCV_CUDA_TRANSFORM_DETAIL_HPP

model_zoo/official/lite/app/src/main/cpp/include/opencv2/core/cuda/detail/type_traits_detail.hpp

@@ -0,0 +1,191 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CUDA_TYPE_TRAITS_DETAIL_HPP
+#define OPENCV_CUDA_TYPE_TRAITS_DETAIL_HPP
+
+#include "../common.hpp"
+#include "../vec_traits.hpp"
+
+//! @cond IGNORED
+
+namespace cv { namespace cuda { namespace device
+{
+    namespace type_traits_detail
+    {
+        template <bool, typename T1, typename T2> struct Select { typedef T1 type; };
+        template <typename T1, typename T2> struct Select<false, T1, T2> { typedef T2 type; };
+
+        template <typename T> struct IsSignedIntergral { enum {value = 0}; };
+        template <> struct IsSignedIntergral<schar> { enum {value = 1}; };
+        template <> struct IsSignedIntergral<char1> { enum {value = 1}; };
+        template <> struct IsSignedIntergral<short> { enum {value = 1}; };
+        template <> struct IsSignedIntergral<short1> { enum {value = 1}; };
+        template <> struct IsSignedIntergral<int> { enum {value = 1}; };
+        template <> struct IsSignedIntergral<int1> { enum {value = 1}; };
+
+        template <typename T> struct IsUnsignedIntegral { enum {value = 0}; };
+        template <> struct IsUnsignedIntegral<uchar> { enum {value = 1}; };
+        template <> struct IsUnsignedIntegral<uchar1> { enum {value = 1}; };
+        template <> struct IsUnsignedIntegral<ushort> { enum {value = 1}; };
+        template <> struct IsUnsignedIntegral<ushort1> { enum {value = 1}; };
+        template <> struct IsUnsignedIntegral<uint> { enum {value = 1}; };
+        template <> struct IsUnsignedIntegral<uint1> { enum {value = 1}; };
+
+        template <typename T> struct IsIntegral { enum {value = IsSignedIntergral<T>::value || IsUnsignedIntegral<T>::value}; };
+        template <> struct IsIntegral<char> { enum {value = 1}; };
+        template <> struct IsIntegral<bool> { enum {value = 1}; };
+
+        template <typename T> struct IsFloat { enum {value = 0}; };
+        template <> struct IsFloat<float> { enum {value = 1}; };
+        template <> struct IsFloat<double> { enum {value = 1}; };
+
+        template <typename T> struct IsVec { enum {value = 0}; };
+        template <> struct IsVec<uchar1> { enum {value = 1}; };
+        template <> struct IsVec<uchar2> { enum {value = 1}; };
+        template <> struct IsVec<uchar3> { enum {value = 1}; };
+        template <> struct IsVec<uchar4> { enum {value = 1}; };
+        template <> struct IsVec<uchar8> { enum {value = 1}; };
+        template <> struct IsVec<char1> { enum {value = 1}; };
+        template <> struct IsVec<char2> { enum {value = 1}; };
+        template <> struct IsVec<char3> { enum {value = 1}; };
+        template <> struct IsVec<char4> { enum {value = 1}; };
+        template <> struct IsVec<char8> { enum {value = 1}; };
+        template <> struct IsVec<ushort1> { enum {value = 1}; };
+        template <> struct IsVec<ushort2> { enum {value = 1}; };
+        template <> struct IsVec<ushort3> { enum {value = 1}; };
+        template <> struct IsVec<ushort4> { enum {value = 1}; };
+        template <> struct IsVec<ushort8> { enum {value = 1}; };
+        template <> struct IsVec<short1> { enum {value = 1}; };
+        template <> struct IsVec<short2> { enum {value = 1}; };
+        template <> struct IsVec<short3> { enum {value = 1}; };
+        template <> struct IsVec<short4> { enum {value = 1}; };
+        template <> struct IsVec<short8> { enum {value = 1}; };
+        template <> struct IsVec<uint1> { enum {value = 1}; };
+        template <> struct IsVec<uint2> { enum {value = 1}; };
+        template <> struct IsVec<uint3> { enum {value = 1}; };
+        template <> struct IsVec<uint4> { enum {value = 1}; };
+        template <> struct IsVec<uint8> { enum {value = 1}; };
+        template <> struct IsVec<int1> { enum {value = 1}; };
+        template <> struct IsVec<int2> { enum {value = 1}; };
+        template <> struct IsVec<int3> { enum {value = 1}; };
+        template <> struct IsVec<int4> { enum {value = 1}; };
+        template <> struct IsVec<int8> { enum {value = 1}; };
+        template <> struct IsVec<float1> { enum {value = 1}; };
+        template <> struct IsVec<float2> { enum {value = 1}; };
+        template <> struct IsVec<float3> { enum {value = 1}; };
+        template <> struct IsVec<float4> { enum {value = 1}; };
+        template <> struct IsVec<float8> { enum {value = 1}; };
+        template <> struct IsVec<double1> { enum {value = 1}; };
+        template <> struct IsVec<double2> { enum {value = 1}; };
+        template <> struct IsVec<double3> { enum {value = 1}; };
+        template <> struct IsVec<double4> { enum {value = 1}; };
+        template <> struct IsVec<double8> { enum {value = 1}; };
+
+        template <class U> struct AddParameterType { typedef const U& type; };
+        template <class U> struct AddParameterType<U&> { typedef U& type; };
+        template <> struct AddParameterType<void> { typedef void type; };
+
+        template <class U> struct ReferenceTraits
+        {
+            enum { value = false };
+            typedef U type;
+        };
+        template <class U> struct ReferenceTraits<U&>
+        {
+            enum { value = true };
+            typedef U type;
+        };
+
+        template <class U> struct PointerTraits
+        {
+            enum { value = false };
+            typedef void type;
+        };
+        template <class U> struct PointerTraits<U*>
+        {
+            enum { value = true };
+            typedef U type;
+        };
+        template <class U> struct PointerTraits<U*&>
+        {
+            enum { value = true };
+            typedef U type;
+        };
+
+        template <class U> struct UnConst
+        {
+            typedef U type;
+            enum { value = 0 };
+        };
+        template <class U> struct UnConst<const U>
+        {
+            typedef U type;
+            enum { value = 1 };
+        };
+        template <class U> struct UnConst<const U&>
+        {
+            typedef U& type;
+            enum { value = 1 };
+        };
+
+        template <class U> struct UnVolatile
+        {
+            typedef U type;
+            enum { value = 0 };
+        };
+        template <class U> struct UnVolatile<volatile U>
+        {
+            typedef U type;
+            enum { value = 1 };
+        };
+        template <class U> struct UnVolatile<volatile U&>
+        {
+            typedef U& type;
+            enum { value = 1 };
+        };
+    } // namespace type_traits_detail
+}}} // namespace cv { namespace cuda { namespace cudev
+
+//! @endcond
+
+#endif // OPENCV_CUDA_TYPE_TRAITS_DETAIL_HPP

model_zoo/official/lite/app/src/main/cpp/include/opencv2/core/cuda/detail/vec_distance_detail.hpp

@@ -0,0 +1,121 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CUDA_VEC_DISTANCE_DETAIL_HPP
+#define OPENCV_CUDA_VEC_DISTANCE_DETAIL_HPP
+
+#include "../datamov_utils.hpp"
+
+//! @cond IGNORED
+
+namespace cv { namespace cuda { namespace device
+{
+    namespace vec_distance_detail
+    {
+        template <int THREAD_DIM, int N> struct UnrollVecDiffCached
+        {
+            template <typename Dist, typename T1, typename T2>
+            static __device__ void calcCheck(const T1* vecCached, const T2* vecGlob, int len, Dist& dist, int ind)
+            {
+                if (ind < len)
+                {
+                    T1 val1 = *vecCached++;
+
+                    T2 val2;
+                    ForceGlob<T2>::Load(vecGlob, ind, val2);
+
+                    dist.reduceIter(val1, val2);
+
+                    UnrollVecDiffCached<THREAD_DIM, N - 1>::calcCheck(vecCached, vecGlob, len, dist, ind + THREAD_DIM);
+                }
+            }
+
+            template <typename Dist, typename T1, typename T2>
+            static __device__ void calcWithoutCheck(const T1* vecCached, const T2* vecGlob, Dist& dist)
+            {
+                T1 val1 = *vecCached++;
+
+                T2 val2;
+                ForceGlob<T2>::Load(vecGlob, 0, val2);
+                vecGlob += THREAD_DIM;
+
+                dist.reduceIter(val1, val2);
+
+                UnrollVecDiffCached<THREAD_DIM, N - 1>::calcWithoutCheck(vecCached, vecGlob, dist);
+            }
+        };
+        template <int THREAD_DIM> struct UnrollVecDiffCached<THREAD_DIM, 0>
+        {
+            template <typename Dist, typename T1, typename T2>
+            static __device__ __forceinline__ void calcCheck(const T1*, const T2*, int, Dist&, int)
+            {
+            }
+
+            template <typename Dist, typename T1, typename T2>
+            static __device__ __forceinline__ void calcWithoutCheck(const T1*, const T2*, Dist&)
+            {
+            }
+        };
+
+        template <int THREAD_DIM, int MAX_LEN, bool LEN_EQ_MAX_LEN> struct VecDiffCachedCalculator;
+        template <int THREAD_DIM, int MAX_LEN> struct VecDiffCachedCalculator<THREAD_DIM, MAX_LEN, false>
+        {
+            template <typename Dist, typename T1, typename T2>
+            static __device__ __forceinline__ void calc(const T1* vecCached, const T2* vecGlob, int len, Dist& dist, int tid)
+            {
+                UnrollVecDiffCached<THREAD_DIM, MAX_LEN / THREAD_DIM>::calcCheck(vecCached, vecGlob, len, dist, tid);
+            }
+        };
+        template <int THREAD_DIM, int MAX_LEN> struct VecDiffCachedCalculator<THREAD_DIM, MAX_LEN, true>
+        {
+            template <typename Dist, typename T1, typename T2>
+            static __device__ __forceinline__ void calc(const T1* vecCached, const T2* vecGlob, int len, Dist& dist, int tid)
+            {
+                UnrollVecDiffCached<THREAD_DIM, MAX_LEN / THREAD_DIM>::calcWithoutCheck(vecCached, vecGlob + tid, dist);
+            }
+        };
+    } // namespace vec_distance_detail
+}}} // namespace cv { namespace cuda { namespace cudev
+
+//! @endcond
+
+#endif // OPENCV_CUDA_VEC_DISTANCE_DETAIL_HPP

model_zoo/official/lite/app/src/main/cpp/include/opencv2/core/cuda/dynamic_smem.hpp

@@ -0,0 +1,88 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CUDA_DYNAMIC_SMEM_HPP
+#define OPENCV_CUDA_DYNAMIC_SMEM_HPP
+
+/** @file
+ * @deprecated Use @ref cudev instead.
+ */
+
+//! @cond IGNORED
+
+namespace cv { namespace cuda { namespace device
+{
+    template<class T> struct DynamicSharedMem
+    {
+        __device__ __forceinline__ operator T*()
+        {
+            extern __shared__ int __smem[];
+            return (T*)__smem;
+        }
+
+        __device__ __forceinline__ operator const T*() const
+        {
+            extern __shared__ int __smem[];
+            return (T*)__smem;
+        }
+    };
+
+    // specialize for double to avoid unaligned memory access compile errors
+    template<> struct DynamicSharedMem<double>
+    {
+        __device__ __forceinline__ operator double*()
+        {
+            extern __shared__ double __smem_d[];
+            return (double*)__smem_d;
+        }
+
+        __device__ __forceinline__ operator const double*() const
+        {
+            extern __shared__ double __smem_d[];
+            return (double*)__smem_d;
+        }
+    };
+}}}
+
+//! @endcond
+
+#endif // OPENCV_CUDA_DYNAMIC_SMEM_HPP

model_zoo/official/lite/app/src/main/cpp/include/opencv2/core/cuda/emulation.hpp

@@ -0,0 +1,269 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CUDA_EMULATION_HPP_
+#define OPENCV_CUDA_EMULATION_HPP_
+
+#include "common.hpp"
+#include "warp_reduce.hpp"
+
+/** @file
+ * @deprecated Use @ref cudev instead.
+ */
+
+//! @cond IGNORED
+
+namespace cv { namespace cuda { namespace device
+{
+    struct Emulation
+    {
+
+        static __device__ __forceinline__ int syncthreadsOr(int pred)
+        {
+#if defined (__CUDA_ARCH__) && (__CUDA_ARCH__ < 200)
+                // just campilation stab
+                return 0;
+#else
+                return __syncthreads_or(pred);
+#endif
+        }
+
+        template<int CTA_SIZE>
+        static __forceinline__ __device__ int Ballot(int predicate)
+        {
+#if defined (__CUDA_ARCH__) && (__CUDA_ARCH__ >= 200)
+            return __ballot(predicate);
+#else
+            __shared__ volatile int cta_buffer[CTA_SIZE];
+
+            int tid = threadIdx.x;
+            cta_buffer[tid] = predicate ? (1 << (tid & 31)) : 0;
+            return warp_reduce(cta_buffer);
+#endif
+        }
+
+        struct smem
+        {
+            enum { TAG_MASK = (1U << ( (sizeof(unsigned int) << 3) - 5U)) - 1U };
+
+            template<typename T>
+            static __device__ __forceinline__ T atomicInc(T* address, T val)
+            {
+#if defined (__CUDA_ARCH__) && (__CUDA_ARCH__ < 120)
+                T count;
+                unsigned int tag = threadIdx.x << ( (sizeof(unsigned int) << 3) - 5U);
+                do
+                {
+                    count = *address & TAG_MASK;
+                    count = tag | (count + 1);
+                    *address = count;
+                } while (*address != count);
+
+                return (count & TAG_MASK) - 1;
+#else
+                return ::atomicInc(address, val);
+#endif
+            }
+
+            template<typename T>
+            static __device__ __forceinline__ T atomicAdd(T* address, T val)
+            {
+#if defined (__CUDA_ARCH__) && (__CUDA_ARCH__ < 120)
+                T count;
+                unsigned int tag = threadIdx.x << ( (sizeof(unsigned int) << 3) - 5U);
+                do
+                {
+                    count = *address & TAG_MASK;
+                    count = tag | (count + val);
+                    *address = count;
+                } while (*address != count);
+
+                return (count & TAG_MASK) - val;
+#else
+                return ::atomicAdd(address, val);
+#endif
+            }
+
+            template<typename T>
+            static __device__ __forceinline__ T atomicMin(T* address, T val)
+            {
+#if defined (__CUDA_ARCH__) && (__CUDA_ARCH__ < 120)
+                T count = ::min(*address, val);
+                do
+                {
+                    *address = count;
+                } while (*address > count);
+
+                return count;
+#else
+                return ::atomicMin(address, val);
+#endif
+            }
+        }; // struct cmem
+
+        struct glob
+        {
+            static __device__ __forceinline__ int atomicAdd(int* address, int val)
+            {
+                return ::atomicAdd(address, val);
+            }
+            static __device__ __forceinline__ unsigned int atomicAdd(unsigned int* address, unsigned int val)
+            {
+                return ::atomicAdd(address, val);
+            }
+            static __device__ __forceinline__ float atomicAdd(float* address, float val)
+            {
+            #if __CUDA_ARCH__ >= 200
+                return ::atomicAdd(address, val);
+            #else
+                int* address_as_i = (int*) address;
+                int old = *address_as_i, assumed;
+                do {
+                    assumed = old;
+                    old = ::atomicCAS(address_as_i, assumed,
+                        __float_as_int(val + __int_as_float(assumed)));
+                } while (assumed != old);
+                return __int_as_float(old);
+            #endif
+            }
+            static __device__ __forceinline__ double atomicAdd(double* address, double val)
+            {
+            #if __CUDA_ARCH__ >= 130
+                unsigned long long int* address_as_ull = (unsigned long long int*) address;
+                unsigned long long int old = *address_as_ull, assumed;
+                do {
+                    assumed = old;
+                    old = ::atomicCAS(address_as_ull, assumed,
+                        __double_as_longlong(val + __longlong_as_double(assumed)));
+                } while (assumed != old);
+                return __longlong_as_double(old);
+            #else
+                CV_UNUSED(address);
+                CV_UNUSED(val);
+                return 0.0;
+            #endif
+            }
+
+            static __device__ __forceinline__ int atomicMin(int* address, int val)
+            {
+                return ::atomicMin(address, val);
+            }
+            static __device__ __forceinline__ float atomicMin(float* address, float val)
+            {
+            #if __CUDA_ARCH__ >= 120
+                int* address_as_i = (int*) address;
+                int old = *address_as_i, assumed;
+                do {
+                    assumed = old;
+                    old = ::atomicCAS(address_as_i, assumed,
+                        __float_as_int(::fminf(val, __int_as_float(assumed))));
+                } while (assumed != old);
+                return __int_as_float(old);
+            #else
+                CV_UNUSED(address);
+                CV_UNUSED(val);
+                return 0.0f;
+            #endif
+            }
+            static __device__ __forceinline__ double atomicMin(double* address, double val)
+            {
+            #if __CUDA_ARCH__ >= 130
+                unsigned long long int* address_as_ull = (unsigned long long int*) address;
+                unsigned long long int old = *address_as_ull, assumed;
+                do {
+                    assumed = old;
+                    old = ::atomicCAS(address_as_ull, assumed,
+                        __double_as_longlong(::fmin(val, __longlong_as_double(assumed))));
+                } while (assumed != old);
+                return __longlong_as_double(old);
+            #else
+                CV_UNUSED(address);
+                CV_UNUSED(val);
+                return 0.0;
+            #endif
+            }
+
+            static __device__ __forceinline__ int atomicMax(int* address, int val)
+            {
+                return ::atomicMax(address, val);
+            }
+            static __device__ __forceinline__ float atomicMax(float* address, float val)
+            {
+            #if __CUDA_ARCH__ >= 120
+                int* address_as_i = (int*) address;
+                int old = *address_as_i, assumed;
+                do {
+                    assumed = old;
+                    old = ::atomicCAS(address_as_i, assumed,
+                        __float_as_int(::fmaxf(val, __int_as_float(assumed))));
+                } while (assumed != old);
+                return __int_as_float(old);
+            #else
+                CV_UNUSED(address);
+                CV_UNUSED(val);
+                return 0.0f;
+            #endif
+            }
+            static __device__ __forceinline__ double atomicMax(double* address, double val)
+            {
+            #if __CUDA_ARCH__ >= 130
+                unsigned long long int* address_as_ull = (unsigned long long int*) address;
+                unsigned long long int old = *address_as_ull, assumed;
+                do {
+                    assumed = old;
+                    old = ::atomicCAS(address_as_ull, assumed,
+                        __double_as_longlong(::fmax(val, __longlong_as_double(assumed))));
+                } while (assumed != old);
+                return __longlong_as_double(old);
+            #else
+                CV_UNUSED(address);
+                CV_UNUSED(val);
+                return 0.0;
+            #endif
+            }
+        };
+    }; //struct Emulation
+}}} // namespace cv { namespace cuda { namespace cudev
+
+//! @endcond
+
+#endif /* OPENCV_CUDA_EMULATION_HPP_ */

model_zoo/official/lite/app/src/main/cpp/include/opencv2/core/cuda/filters.hpp

@@ -0,0 +1,286 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CUDA_FILTERS_HPP
+#define OPENCV_CUDA_FILTERS_HPP
+
+#include "saturate_cast.hpp"
+#include "vec_traits.hpp"
+#include "vec_math.hpp"
+#include "type_traits.hpp"
+
+/** @file
+ * @deprecated Use @ref cudev instead.
+ */
+
+//! @cond IGNORED
+
+namespace cv { namespace cuda { namespace device
+{
+    template <typename Ptr2D> struct PointFilter
+    {
+        typedef typename Ptr2D::elem_type elem_type;
+        typedef float index_type;
+
+        explicit __host__ __device__ __forceinline__ PointFilter(const Ptr2D& src_, float fx = 0.f, float fy = 0.f)
+        : src(src_)
+        {
+            CV_UNUSED(fx);
+            CV_UNUSED(fy);
+        }
+
+        __device__ __forceinline__ elem_type operator ()(float y, float x) const
+        {
+            return src(__float2int_rz(y), __float2int_rz(x));
+        }
+
+        Ptr2D src;
+    };
+
+    template <typename Ptr2D> struct LinearFilter
+    {
+        typedef typename Ptr2D::elem_type elem_type;
+        typedef float index_type;
+
+        explicit __host__ __device__ __forceinline__ LinearFilter(const Ptr2D& src_, float fx = 0.f, float fy = 0.f)
+        : src(src_)
+        {
+            CV_UNUSED(fx);
+            CV_UNUSED(fy);
+        }
+        __device__ __forceinline__ elem_type operator ()(float y, float x) const
+        {
+            typedef typename TypeVec<float, VecTraits<elem_type>::cn>::vec_type work_type;
+
+            work_type out = VecTraits<work_type>::all(0);
+
+            const int x1 = __float2int_rd(x);
+            const int y1 = __float2int_rd(y);
+            const int x2 = x1 + 1;
+            const int y2 = y1 + 1;
+
+            elem_type src_reg = src(y1, x1);
+            out = out + src_reg * ((x2 - x) * (y2 - y));
+
+            src_reg = src(y1, x2);
+            out = out + src_reg * ((x - x1) * (y2 - y));
+
+            src_reg = src(y2, x1);
+            out = out + src_reg * ((x2 - x) * (y - y1));
+
+            src_reg = src(y2, x2);
+            out = out + src_reg * ((x - x1) * (y - y1));
+
+            return saturate_cast<elem_type>(out);
+        }
+
+        Ptr2D src;
+    };
+
+    template <typename Ptr2D> struct CubicFilter
+    {
+        typedef typename Ptr2D::elem_type elem_type;
+        typedef float index_type;
+        typedef typename TypeVec<float, VecTraits<elem_type>::cn>::vec_type work_type;
+
+        explicit __host__ __device__ __forceinline__ CubicFilter(const Ptr2D& src_, float fx = 0.f, float fy = 0.f)
+        : src(src_)
+        {
+            CV_UNUSED(fx);
+            CV_UNUSED(fy);
+        }
+
+        static __device__ __forceinline__ float bicubicCoeff(float x_)
+        {
+            float x = fabsf(x_);
+            if (x <= 1.0f)
+            {
+                return x * x * (1.5f * x - 2.5f) + 1.0f;
+            }
+            else if (x < 2.0f)
+            {
+                return x * (x * (-0.5f * x + 2.5f) - 4.0f) + 2.0f;
+            }
+            else
+            {
+                return 0.0f;
+            }
+        }
+
+        __device__ elem_type operator ()(float y, float x) const
+        {
+            const float xmin = ::ceilf(x - 2.0f);
+            const float xmax = ::floorf(x + 2.0f);
+
+            const float ymin = ::ceilf(y - 2.0f);
+            const float ymax = ::floorf(y + 2.0f);
+
+            work_type sum = VecTraits<work_type>::all(0);
+            float wsum = 0.0f;
+
+            for (float cy = ymin; cy <= ymax; cy += 1.0f)
+            {
+                for (float cx = xmin; cx <= xmax; cx += 1.0f)
+                {
+                    const float w = bicubicCoeff(x - cx) * bicubicCoeff(y - cy);
+                    sum = sum + w * src(__float2int_rd(cy), __float2int_rd(cx));
+                    wsum += w;
+                }
+            }
+
+            work_type res = (!wsum)? VecTraits<work_type>::all(0) : sum / wsum;
+
+            return saturate_cast<elem_type>(res);
+        }
+
+        Ptr2D src;
+    };
+    // for integer scaling
+    template <typename Ptr2D> struct IntegerAreaFilter
+    {
+        typedef typename Ptr2D::elem_type elem_type;
+        typedef float index_type;
+
+        explicit __host__ __device__ __forceinline__ IntegerAreaFilter(const Ptr2D& src_, float scale_x_, float scale_y_)
+            : src(src_), scale_x(scale_x_), scale_y(scale_y_), scale(1.f / (scale_x * scale_y)) {}
+
+        __device__ __forceinline__ elem_type operator ()(float y, float x) const
+        {
+            float fsx1 = x * scale_x;
+            float fsx2 = fsx1 + scale_x;
+
+            int sx1 = __float2int_ru(fsx1);
+            int sx2 = __float2int_rd(fsx2);
+
+            float fsy1 = y * scale_y;
+            float fsy2 = fsy1 + scale_y;
+
+            int sy1 = __float2int_ru(fsy1);
+            int sy2 = __float2int_rd(fsy2);
+
+            typedef typename TypeVec<float, VecTraits<elem_type>::cn>::vec_type work_type;
+            work_type out = VecTraits<work_type>::all(0.f);
+
+            for(int dy = sy1; dy < sy2; ++dy)
+                for(int dx = sx1; dx < sx2; ++dx)
+                {
+                    out = out + src(dy, dx) * scale;
+                }
+
+            return saturate_cast<elem_type>(out);
+        }
+
+        Ptr2D src;
+        float scale_x, scale_y ,scale;
+    };
+
+    template <typename Ptr2D> struct AreaFilter
+    {
+        typedef typename Ptr2D::elem_type elem_type;
+        typedef float index_type;
+
+        explicit __host__ __device__ __forceinline__ AreaFilter(const Ptr2D& src_, float scale_x_, float scale_y_)
+            : src(src_), scale_x(scale_x_), scale_y(scale_y_){}
+
+        __device__ __forceinline__ elem_type operator ()(float y, float x) const
+        {
+            float fsx1 = x * scale_x;
+            float fsx2 = fsx1 + scale_x;
+
+            int sx1 = __float2int_ru(fsx1);
+            int sx2 = __float2int_rd(fsx2);
+
+            float fsy1 = y * scale_y;
+            float fsy2 = fsy1 + scale_y;
+
+            int sy1 = __float2int_ru(fsy1);
+            int sy2 = __float2int_rd(fsy2);
+
+            float scale = 1.f / (fminf(scale_x, src.width - fsx1) * fminf(scale_y, src.height - fsy1));
+
+            typedef typename TypeVec<float, VecTraits<elem_type>::cn>::vec_type work_type;
+            work_type out = VecTraits<work_type>::all(0.f);
+
+            for (int dy = sy1; dy < sy2; ++dy)
+            {
+                for (int dx = sx1; dx < sx2; ++dx)
+                    out = out + src(dy, dx) * scale;
+
+                if (sx1 > fsx1)
+                    out = out + src(dy, (sx1 -1) ) * ((sx1 - fsx1) * scale);
+
+                if (sx2 < fsx2)
+                    out = out + src(dy, sx2) * ((fsx2 -sx2) * scale);
+            }
+
+            if (sy1 > fsy1)
+                for (int dx = sx1; dx < sx2; ++dx)
+                    out = out + src( (sy1 - 1) , dx) * ((sy1 -fsy1) * scale);
+
+            if (sy2 < fsy2)
+                for (int dx = sx1; dx < sx2; ++dx)
+                    out = out + src(sy2, dx) * ((fsy2 -sy2) * scale);
+
+            if ((sy1 > fsy1) &&  (sx1 > fsx1))
+                out = out + src( (sy1 - 1) , (sx1 - 1)) * ((sy1 -fsy1) * (sx1 -fsx1) * scale);
+
+            if ((sy1 > fsy1) &&  (sx2 < fsx2))
+                out = out + src( (sy1 - 1) , sx2) * ((sy1 -fsy1) * (fsx2 -sx2) * scale);
+
+            if ((sy2 < fsy2) &&  (sx2 < fsx2))
+                out = out + src(sy2, sx2) * ((fsy2 -sy2) * (fsx2 -sx2) * scale);
+
+            if ((sy2 < fsy2) &&  (sx1 > fsx1))
+                out = out + src(sy2, (sx1 - 1)) * ((fsy2 -sy2) * (sx1 -fsx1) * scale);
+
+            return saturate_cast<elem_type>(out);
+        }
+
+        Ptr2D src;
+        float scale_x, scale_y;
+        int width, haight;
+    };
+}}} // namespace cv { namespace cuda { namespace cudev
+
+//! @endcond
+
+#endif // OPENCV_CUDA_FILTERS_HPP

model_zoo/official/lite/app/src/main/cpp/include/opencv2/core/cuda/funcattrib.hpp

@@ -0,0 +1,79 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CUDA_DEVICE_FUNCATTRIB_HPP
+#define OPENCV_CUDA_DEVICE_FUNCATTRIB_HPP
+
+#include <cstdio>
+
+/** @file
+ * @deprecated Use @ref cudev instead.
+ */
+
+//! @cond IGNORED
+
+namespace cv { namespace cuda { namespace device
+{
+    template<class Func>
+    void printFuncAttrib(Func& func)
+    {
+
+        cudaFuncAttributes attrs;
+        cudaFuncGetAttributes(&attrs, func);
+
+        printf("=== Function stats ===\n");
+        printf("Name: \n");
+        printf("sharedSizeBytes    = %d\n", attrs.sharedSizeBytes);
+        printf("constSizeBytes     = %d\n", attrs.constSizeBytes);
+        printf("localSizeBytes     = %d\n", attrs.localSizeBytes);
+        printf("maxThreadsPerBlock = %d\n", attrs.maxThreadsPerBlock);
+        printf("numRegs            = %d\n", attrs.numRegs);
+        printf("ptxVersion         = %d\n", attrs.ptxVersion);
+        printf("binaryVersion      = %d\n", attrs.binaryVersion);
+        printf("\n");
+        fflush(stdout);
+    }
+}}} // namespace cv { namespace cuda { namespace cudev
+
+//! @endcond
+
+#endif  /* OPENCV_CUDA_DEVICE_FUNCATTRIB_HPP */

model_zoo/official/lite/app/src/main/cpp/include/opencv2/core/cuda/functional.hpp

@@ -0,0 +1,805 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CUDA_FUNCTIONAL_HPP
+#define OPENCV_CUDA_FUNCTIONAL_HPP
+
+#include <functional>
+#include "saturate_cast.hpp"
+#include "vec_traits.hpp"
+#include "type_traits.hpp"
+
+/** @file
+ * @deprecated Use @ref cudev instead.
+ */
+
+//! @cond IGNORED
+
+namespace cv { namespace cuda { namespace device
+{
+    // Function Objects
+    template<typename Argument, typename Result> struct unary_function
+    {
+        typedef Argument argument_type;
+        typedef Result result_type;
+    };
+    template<typename Argument1, typename Argument2, typename Result> struct binary_function
+    {
+        typedef Argument1 first_argument_type;
+        typedef Argument2 second_argument_type;
+        typedef Result result_type;
+    };
+
+    // Arithmetic Operations
+    template <typename T> struct plus : binary_function<T, T, T>
+    {
+        __device__ __forceinline__ T operator ()(typename TypeTraits<T>::ParameterType a,
+                                                 typename TypeTraits<T>::ParameterType b) const
+        {
+            return a + b;
+        }
+        __host__ __device__ __forceinline__ plus() {}
+        __host__ __device__ __forceinline__ plus(const plus&) {}
+    };
+
+    template <typename T> struct minus : binary_function<T, T, T>
+    {
+        __device__ __forceinline__ T operator ()(typename TypeTraits<T>::ParameterType a,
+                                                 typename TypeTraits<T>::ParameterType b) const
+        {
+            return a - b;
+        }
+        __host__ __device__ __forceinline__ minus() {}
+        __host__ __device__ __forceinline__ minus(const minus&) {}
+    };
+
+    template <typename T> struct multiplies : binary_function<T, T, T>
+    {
+        __device__ __forceinline__ T operator ()(typename TypeTraits<T>::ParameterType a,
+                                                 typename TypeTraits<T>::ParameterType b) const
+        {
+            return a * b;
+        }
+        __host__ __device__ __forceinline__ multiplies() {}
+        __host__ __device__ __forceinline__ multiplies(const multiplies&) {}
+    };
+
+    template <typename T> struct divides : binary_function<T, T, T>
+    {
+        __device__ __forceinline__ T operator ()(typename TypeTraits<T>::ParameterType a,
+                                                 typename TypeTraits<T>::ParameterType b) const
+        {
+            return a / b;
+        }
+        __host__ __device__ __forceinline__ divides() {}
+        __host__ __device__ __forceinline__ divides(const divides&) {}
+    };
+
+    template <typename T> struct modulus : binary_function<T, T, T>
+    {
+        __device__ __forceinline__ T operator ()(typename TypeTraits<T>::ParameterType a,
+                                                 typename TypeTraits<T>::ParameterType b) const
+        {
+            return a % b;
+        }
+        __host__ __device__ __forceinline__ modulus() {}
+        __host__ __device__ __forceinline__ modulus(const modulus&) {}
+    };
+
+    template <typename T> struct negate : unary_function<T, T>
+    {
+        __device__ __forceinline__ T operator ()(typename TypeTraits<T>::ParameterType a) const
+        {
+            return -a;
+        }
+        __host__ __device__ __forceinline__ negate() {}
+        __host__ __device__ __forceinline__ negate(const negate&) {}
+    };
+
+    // Comparison Operations
+    template <typename T> struct equal_to : binary_function<T, T, bool>
+    {
+        __device__ __forceinline__ bool operator ()(typename TypeTraits<T>::ParameterType a,
+                                                    typename TypeTraits<T>::ParameterType b) const
+        {
+            return a == b;
+        }
+        __host__ __device__ __forceinline__ equal_to() {}
+        __host__ __device__ __forceinline__ equal_to(const equal_to&) {}
+    };
+
+    template <typename T> struct not_equal_to : binary_function<T, T, bool>
+    {
+        __device__ __forceinline__ bool operator ()(typename TypeTraits<T>::ParameterType a,
+                                                    typename TypeTraits<T>::ParameterType b) const
+        {
+            return a != b;
+        }
+        __host__ __device__ __forceinline__ not_equal_to() {}
+        __host__ __device__ __forceinline__ not_equal_to(const not_equal_to&) {}
+    };
+
+    template <typename T> struct greater : binary_function<T, T, bool>
+    {
+        __device__ __forceinline__ bool operator ()(typename TypeTraits<T>::ParameterType a,
+                                                    typename TypeTraits<T>::ParameterType b) const
+        {
+            return a > b;
+        }
+        __host__ __device__ __forceinline__ greater() {}
+        __host__ __device__ __forceinline__ greater(const greater&) {}
+    };
+
+    template <typename T> struct less : binary_function<T, T, bool>
+    {
+        __device__ __forceinline__ bool operator ()(typename TypeTraits<T>::ParameterType a,
+                                                    typename TypeTraits<T>::ParameterType b) const
+        {
+            return a < b;
+        }
+        __host__ __device__ __forceinline__ less() {}
+        __host__ __device__ __forceinline__ less(const less&) {}
+    };
+
+    template <typename T> struct greater_equal : binary_function<T, T, bool>
+    {
+        __device__ __forceinline__ bool operator ()(typename TypeTraits<T>::ParameterType a,
+                                                    typename TypeTraits<T>::ParameterType b) const
+        {
+            return a >= b;
+        }
+        __host__ __device__ __forceinline__ greater_equal() {}
+        __host__ __device__ __forceinline__ greater_equal(const greater_equal&) {}
+    };
+
+    template <typename T> struct less_equal : binary_function<T, T, bool>
+    {
+        __device__ __forceinline__ bool operator ()(typename TypeTraits<T>::ParameterType a,
+                                                    typename TypeTraits<T>::ParameterType b) const
+        {
+            return a <= b;
+        }
+        __host__ __device__ __forceinline__ less_equal() {}
+        __host__ __device__ __forceinline__ less_equal(const less_equal&) {}
+    };
+
+    // Logical Operations
+    template <typename T> struct logical_and : binary_function<T, T, bool>
+    {
+        __device__ __forceinline__ bool operator ()(typename TypeTraits<T>::ParameterType a,
+                                                    typename TypeTraits<T>::ParameterType b) const
+        {
+            return a && b;
+        }
+        __host__ __device__ __forceinline__ logical_and() {}
+        __host__ __device__ __forceinline__ logical_and(const logical_and&) {}
+    };
+
+    template <typename T> struct logical_or : binary_function<T, T, bool>
+    {
+        __device__ __forceinline__ bool operator ()(typename TypeTraits<T>::ParameterType a,
+                                                    typename TypeTraits<T>::ParameterType b) const
+        {
+            return a || b;
+        }
+        __host__ __device__ __forceinline__ logical_or() {}
+        __host__ __device__ __forceinline__ logical_or(const logical_or&) {}
+    };
+
+    template <typename T> struct logical_not : unary_function<T, bool>
+    {
+        __device__ __forceinline__ bool operator ()(typename TypeTraits<T>::ParameterType a) const
+        {
+            return !a;
+        }
+        __host__ __device__ __forceinline__ logical_not() {}
+        __host__ __device__ __forceinline__ logical_not(const logical_not&) {}
+    };
+
+    // Bitwise Operations
+    template <typename T> struct bit_and : binary_function<T, T, T>
+    {
+        __device__ __forceinline__ T operator ()(typename TypeTraits<T>::ParameterType a,
+                                                 typename TypeTraits<T>::ParameterType b) const
+        {
+            return a & b;
+        }
+        __host__ __device__ __forceinline__ bit_and() {}
+        __host__ __device__ __forceinline__ bit_and(const bit_and&) {}
+    };
+
+    template <typename T> struct bit_or : binary_function<T, T, T>
+    {
+        __device__ __forceinline__ T operator ()(typename TypeTraits<T>::ParameterType a,
+                                                 typename TypeTraits<T>::ParameterType b) const
+        {
+            return a | b;
+        }
+        __host__ __device__ __forceinline__ bit_or() {}
+        __host__ __device__ __forceinline__ bit_or(const bit_or&) {}
+    };
+
+    template <typename T> struct bit_xor : binary_function<T, T, T>
+    {
+        __device__ __forceinline__ T operator ()(typename TypeTraits<T>::ParameterType a,
+                                                 typename TypeTraits<T>::ParameterType b) const
+        {
+            return a ^ b;
+        }
+        __host__ __device__ __forceinline__ bit_xor() {}
+        __host__ __device__ __forceinline__ bit_xor(const bit_xor&) {}
+    };
+
+    template <typename T> struct bit_not : unary_function<T, T>
+    {
+        __device__ __forceinline__ T operator ()(typename TypeTraits<T>::ParameterType v) const
+        {
+            return ~v;
+        }
+        __host__ __device__ __forceinline__ bit_not() {}
+        __host__ __device__ __forceinline__ bit_not(const bit_not&) {}
+    };
+
+    // Generalized Identity Operations
+    template <typename T> struct identity : unary_function<T, T>
+    {
+        __device__ __forceinline__ typename TypeTraits<T>::ParameterType operator()(typename TypeTraits<T>::ParameterType x) const
+        {
+            return x;
+        }
+        __host__ __device__ __forceinline__ identity() {}
+        __host__ __device__ __forceinline__ identity(const identity&) {}
+    };
+
+    template <typename T1, typename T2> struct project1st : binary_function<T1, T2, T1>
+    {
+        __device__ __forceinline__ typename TypeTraits<T1>::ParameterType operator()(typename TypeTraits<T1>::ParameterType lhs, typename TypeTraits<T2>::ParameterType rhs) const
+        {
+            return lhs;
+        }
+        __host__ __device__ __forceinline__ project1st() {}
+        __host__ __device__ __forceinline__ project1st(const project1st&) {}
+    };
+
+    template <typename T1, typename T2> struct project2nd : binary_function<T1, T2, T2>
+    {
+        __device__ __forceinline__ typename TypeTraits<T2>::ParameterType operator()(typename TypeTraits<T1>::ParameterType lhs, typename TypeTraits<T2>::ParameterType rhs) const
+        {
+            return rhs;
+        }
+        __host__ __device__ __forceinline__ project2nd() {}
+        __host__ __device__ __forceinline__ project2nd(const project2nd&) {}
+    };
+
+    // Min/Max Operations
+
+#define OPENCV_CUDA_IMPLEMENT_MINMAX(name, type, op) \
+    template <> struct name<type> : binary_function<type, type, type> \
+    { \
+        __device__ __forceinline__ type operator()(type lhs, type rhs) const {return op(lhs, rhs);} \
+        __host__ __device__ __forceinline__ name() {}\
+        __host__ __device__ __forceinline__ name(const name&) {}\
+    };
+
+    template <typename T> struct maximum : binary_function<T, T, T>
+    {
+        __device__ __forceinline__ T operator()(typename TypeTraits<T>::ParameterType lhs, typename TypeTraits<T>::ParameterType rhs) const
+        {
+            return max(lhs, rhs);
+        }
+        __host__ __device__ __forceinline__ maximum() {}
+        __host__ __device__ __forceinline__ maximum(const maximum&) {}
+    };
+
+    OPENCV_CUDA_IMPLEMENT_MINMAX(maximum, uchar, ::max)
+    OPENCV_CUDA_IMPLEMENT_MINMAX(maximum, schar, ::max)
+    OPENCV_CUDA_IMPLEMENT_MINMAX(maximum, char, ::max)
+    OPENCV_CUDA_IMPLEMENT_MINMAX(maximum, ushort, ::max)
+    OPENCV_CUDA_IMPLEMENT_MINMAX(maximum, short, ::max)
+    OPENCV_CUDA_IMPLEMENT_MINMAX(maximum, int, ::max)
+    OPENCV_CUDA_IMPLEMENT_MINMAX(maximum, uint, ::max)
+    OPENCV_CUDA_IMPLEMENT_MINMAX(maximum, float, ::fmax)
+    OPENCV_CUDA_IMPLEMENT_MINMAX(maximum, double, ::fmax)
+
+    template <typename T> struct minimum : binary_function<T, T, T>
+    {
+        __device__ __forceinline__ T operator()(typename TypeTraits<T>::ParameterType lhs, typename TypeTraits<T>::ParameterType rhs) const
+        {
+            return min(lhs, rhs);
+        }
+        __host__ __device__ __forceinline__ minimum() {}
+        __host__ __device__ __forceinline__ minimum(const minimum&) {}
+    };
+
+    OPENCV_CUDA_IMPLEMENT_MINMAX(minimum, uchar, ::min)
+    OPENCV_CUDA_IMPLEMENT_MINMAX(minimum, schar, ::min)
+    OPENCV_CUDA_IMPLEMENT_MINMAX(minimum, char, ::min)
+    OPENCV_CUDA_IMPLEMENT_MINMAX(minimum, ushort, ::min)
+    OPENCV_CUDA_IMPLEMENT_MINMAX(minimum, short, ::min)
+    OPENCV_CUDA_IMPLEMENT_MINMAX(minimum, int, ::min)
+    OPENCV_CUDA_IMPLEMENT_MINMAX(minimum, uint, ::min)
+    OPENCV_CUDA_IMPLEMENT_MINMAX(minimum, float, ::fmin)
+    OPENCV_CUDA_IMPLEMENT_MINMAX(minimum, double, ::fmin)
+
+#undef OPENCV_CUDA_IMPLEMENT_MINMAX
+
+    // Math functions
+
+    template <typename T> struct abs_func : unary_function<T, T>
+    {
+        __device__ __forceinline__ T operator ()(typename TypeTraits<T>::ParameterType x) const
+        {
+            return abs(x);
+        }
+
+        __host__ __device__ __forceinline__ abs_func() {}
+        __host__ __device__ __forceinline__ abs_func(const abs_func&) {}
+    };
+    template <> struct abs_func<unsigned char> : unary_function<unsigned char, unsigned char>
+    {
+        __device__ __forceinline__ unsigned char operator ()(unsigned char x) const
+        {
+            return x;
+        }
+
+        __host__ __device__ __forceinline__ abs_func() {}
+        __host__ __device__ __forceinline__ abs_func(const abs_func&) {}
+    };
+    template <> struct abs_func<signed char> : unary_function<signed char, signed char>
+    {
+        __device__ __forceinline__ signed char operator ()(signed char x) const
+        {
+            return ::abs((int)x);
+        }
+
+        __host__ __device__ __forceinline__ abs_func() {}
+        __host__ __device__ __forceinline__ abs_func(const abs_func&) {}
+    };
+    template <> struct abs_func<char> : unary_function<char, char>
+    {
+        __device__ __forceinline__ char operator ()(char x) const
+        {
+            return ::abs((int)x);
+        }
+
+        __host__ __device__ __forceinline__ abs_func() {}
+        __host__ __device__ __forceinline__ abs_func(const abs_func&) {}
+    };
+    template <> struct abs_func<unsigned short> : unary_function<unsigned short, unsigned short>
+    {
+        __device__ __forceinline__ unsigned short operator ()(unsigned short x) const
+        {
+            return x;
+        }
+
+        __host__ __device__ __forceinline__ abs_func() {}
+        __host__ __device__ __forceinline__ abs_func(const abs_func&) {}
+    };
+    template <> struct abs_func<short> : unary_function<short, short>
+    {
+        __device__ __forceinline__ short operator ()(short x) const
+        {
+            return ::abs((int)x);
+        }
+
+        __host__ __device__ __forceinline__ abs_func() {}
+        __host__ __device__ __forceinline__ abs_func(const abs_func&) {}
+    };
+    template <> struct abs_func<unsigned int> : unary_function<unsigned int, unsigned int>
+    {
+        __device__ __forceinline__ unsigned int operator ()(unsigned int x) const
+        {
+            return x;
+        }
+
+        __host__ __device__ __forceinline__ abs_func() {}
+        __host__ __device__ __forceinline__ abs_func(const abs_func&) {}
+    };
+    template <> struct abs_func<int> : unary_function<int, int>
+    {
+        __device__ __forceinline__ int operator ()(int x) const
+        {
+            return ::abs(x);
+        }
+
+        __host__ __device__ __forceinline__ abs_func() {}
+        __host__ __device__ __forceinline__ abs_func(const abs_func&) {}
+    };
+    template <> struct abs_func<float> : unary_function<float, float>
+    {
+        __device__ __forceinline__ float operator ()(float x) const
+        {
+            return ::fabsf(x);
+        }
+
+        __host__ __device__ __forceinline__ abs_func() {}
+        __host__ __device__ __forceinline__ abs_func(const abs_func&) {}
+    };
+    template <> struct abs_func<double> : unary_function<double, double>
+    {
+        __device__ __forceinline__ double operator ()(double x) const
+        {
+            return ::fabs(x);
+        }
+
+        __host__ __device__ __forceinline__ abs_func() {}
+        __host__ __device__ __forceinline__ abs_func(const abs_func&) {}
+    };
+
+#define OPENCV_CUDA_IMPLEMENT_UN_FUNCTOR(name, func) \
+    template <typename T> struct name ## _func : unary_function<T, float> \
+    { \
+        __device__ __forceinline__ float operator ()(typename TypeTraits<T>::ParameterType v) const \
+        { \
+            return func ## f(v); \
+        } \
+        __host__ __device__ __forceinline__ name ## _func() {} \
+        __host__ __device__ __forceinline__ name ## _func(const name ## _func&) {} \
+    }; \
+    template <> struct name ## _func<double> : unary_function<double, double> \
+    { \
+        __device__ __forceinline__ double operator ()(double v) const \
+        { \
+            return func(v); \
+        } \
+        __host__ __device__ __forceinline__ name ## _func() {} \
+        __host__ __device__ __forceinline__ name ## _func(const name ## _func&) {} \
+    };
+
+#define OPENCV_CUDA_IMPLEMENT_BIN_FUNCTOR(name, func) \
+    template <typename T> struct name ## _func : binary_function<T, T, float> \
+    { \
+        __device__ __forceinline__ float operator ()(typename TypeTraits<T>::ParameterType v1, typename TypeTraits<T>::ParameterType v2) const \
+        { \
+            return func ## f(v1, v2); \
+        } \
+        __host__ __device__ __forceinline__ name ## _func() {} \
+        __host__ __device__ __forceinline__ name ## _func(const name ## _func&) {} \
+    }; \
+    template <> struct name ## _func<double> : binary_function<double, double, double> \
+    { \
+        __device__ __forceinline__ double operator ()(double v1, double v2) const \
+        { \
+            return func(v1, v2); \
+        } \
+        __host__ __device__ __forceinline__ name ## _func() {} \
+        __host__ __device__ __forceinline__ name ## _func(const name ## _func&) {} \
+    };
+
+    OPENCV_CUDA_IMPLEMENT_UN_FUNCTOR(sqrt, ::sqrt)
+    OPENCV_CUDA_IMPLEMENT_UN_FUNCTOR(exp, ::exp)
+    OPENCV_CUDA_IMPLEMENT_UN_FUNCTOR(exp2, ::exp2)
+    OPENCV_CUDA_IMPLEMENT_UN_FUNCTOR(exp10, ::exp10)
+    OPENCV_CUDA_IMPLEMENT_UN_FUNCTOR(log, ::log)
+    OPENCV_CUDA_IMPLEMENT_UN_FUNCTOR(log2, ::log2)
+    OPENCV_CUDA_IMPLEMENT_UN_FUNCTOR(log10, ::log10)
+    OPENCV_CUDA_IMPLEMENT_UN_FUNCTOR(sin, ::sin)
+    OPENCV_CUDA_IMPLEMENT_UN_FUNCTOR(cos, ::cos)
+    OPENCV_CUDA_IMPLEMENT_UN_FUNCTOR(tan, ::tan)
+    OPENCV_CUDA_IMPLEMENT_UN_FUNCTOR(asin, ::asin)
+    OPENCV_CUDA_IMPLEMENT_UN_FUNCTOR(acos, ::acos)
+    OPENCV_CUDA_IMPLEMENT_UN_FUNCTOR(atan, ::atan)
+    OPENCV_CUDA_IMPLEMENT_UN_FUNCTOR(sinh, ::sinh)
+    OPENCV_CUDA_IMPLEMENT_UN_FUNCTOR(cosh, ::cosh)
+    OPENCV_CUDA_IMPLEMENT_UN_FUNCTOR(tanh, ::tanh)
+    OPENCV_CUDA_IMPLEMENT_UN_FUNCTOR(asinh, ::asinh)
+    OPENCV_CUDA_IMPLEMENT_UN_FUNCTOR(acosh, ::acosh)
+    OPENCV_CUDA_IMPLEMENT_UN_FUNCTOR(atanh, ::atanh)
+
+    OPENCV_CUDA_IMPLEMENT_BIN_FUNCTOR(hypot, ::hypot)
+    OPENCV_CUDA_IMPLEMENT_BIN_FUNCTOR(atan2, ::atan2)
+    OPENCV_CUDA_IMPLEMENT_BIN_FUNCTOR(pow, ::pow)
+
+    #undef OPENCV_CUDA_IMPLEMENT_UN_FUNCTOR
+    #undef OPENCV_CUDA_IMPLEMENT_UN_FUNCTOR_NO_DOUBLE
+    #undef OPENCV_CUDA_IMPLEMENT_BIN_FUNCTOR
+
+    template<typename T> struct hypot_sqr_func : binary_function<T, T, float>
+    {
+        __device__ __forceinline__ T operator ()(typename TypeTraits<T>::ParameterType src1, typename TypeTraits<T>::ParameterType src2) const
+        {
+            return src1 * src1 + src2 * src2;
+        }
+        __host__ __device__ __forceinline__ hypot_sqr_func() {}
+        __host__ __device__ __forceinline__ hypot_sqr_func(const hypot_sqr_func&) {}
+    };
+
+    // Saturate Cast Functor
+    template <typename T, typename D> struct saturate_cast_func : unary_function<T, D>
+    {
+        __device__ __forceinline__ D operator ()(typename TypeTraits<T>::ParameterType v) const
+        {
+            return saturate_cast<D>(v);
+        }
+        __host__ __device__ __forceinline__ saturate_cast_func() {}
+        __host__ __device__ __forceinline__ saturate_cast_func(const saturate_cast_func&) {}
+    };
+
+    // Threshold Functors
+    template <typename T> struct thresh_binary_func : unary_function<T, T>
+    {
+        __host__ __device__ __forceinline__ thresh_binary_func(T thresh_, T maxVal_) : thresh(thresh_), maxVal(maxVal_) {}
+
+        __device__ __forceinline__ T operator()(typename TypeTraits<T>::ParameterType src) const
+        {
+            return (src > thresh) * maxVal;
+        }
+
+        __host__ __device__ __forceinline__ thresh_binary_func() {}
+        __host__ __device__ __forceinline__ thresh_binary_func(const thresh_binary_func& other)
+            : thresh(other.thresh), maxVal(other.maxVal) {}
+
+        T thresh;
+        T maxVal;
+    };
+
+    template <typename T> struct thresh_binary_inv_func : unary_function<T, T>
+    {
+        __host__ __device__ __forceinline__ thresh_binary_inv_func(T thresh_, T maxVal_) : thresh(thresh_), maxVal(maxVal_) {}
+
+        __device__ __forceinline__ T operator()(typename TypeTraits<T>::ParameterType src) const
+        {
+            return (src <= thresh) * maxVal;
+        }
+
+        __host__ __device__ __forceinline__ thresh_binary_inv_func() {}
+        __host__ __device__ __forceinline__ thresh_binary_inv_func(const thresh_binary_inv_func& other)
+            : thresh(other.thresh), maxVal(other.maxVal) {}
+
+        T thresh;
+        T maxVal;
+    };
+
+    template <typename T> struct thresh_trunc_func : unary_function<T, T>
+    {
+        explicit __host__ __device__ __forceinline__ thresh_trunc_func(T thresh_, T maxVal_ = 0) : thresh(thresh_) {CV_UNUSED(maxVal_);}
+
+        __device__ __forceinline__ T operator()(typename TypeTraits<T>::ParameterType src) const
+        {
+            return minimum<T>()(src, thresh);
+        }
+
+        __host__ __device__ __forceinline__ thresh_trunc_func() {}
+        __host__ __device__ __forceinline__ thresh_trunc_func(const thresh_trunc_func& other)
+            : thresh(other.thresh) {}
+
+        T thresh;
+    };
+
+    template <typename T> struct thresh_to_zero_func : unary_function<T, T>
+    {
+        explicit __host__ __device__ __forceinline__ thresh_to_zero_func(T thresh_, T maxVal_ = 0) : thresh(thresh_) {CV_UNUSED(maxVal_);}
+
+        __device__ __forceinline__ T operator()(typename TypeTraits<T>::ParameterType src) const
+        {
+            return (src > thresh) * src;
+        }
+
+        __host__ __device__ __forceinline__ thresh_to_zero_func() {}
+       __host__  __device__ __forceinline__ thresh_to_zero_func(const thresh_to_zero_func& other)
+            : thresh(other.thresh) {}
+
+        T thresh;
+    };
+
+    template <typename T> struct thresh_to_zero_inv_func : unary_function<T, T>
+    {
+        explicit __host__ __device__ __forceinline__ thresh_to_zero_inv_func(T thresh_, T maxVal_ = 0) : thresh(thresh_) {CV_UNUSED(maxVal_);}
+
+        __device__ __forceinline__ T operator()(typename TypeTraits<T>::ParameterType src) const
+        {
+            return (src <= thresh) * src;
+        }
+
+        __host__ __device__ __forceinline__ thresh_to_zero_inv_func() {}
+        __host__ __device__ __forceinline__ thresh_to_zero_inv_func(const thresh_to_zero_inv_func& other)
+            : thresh(other.thresh) {}
+
+        T thresh;
+    };
+
+    // Function Object Adaptors
+    template <typename Predicate> struct unary_negate : unary_function<typename Predicate::argument_type, bool>
+    {
+      explicit __host__ __device__ __forceinline__ unary_negate(const Predicate& p) : pred(p) {}
+
+      __device__ __forceinline__ bool operator()(typename TypeTraits<typename Predicate::argument_type>::ParameterType x) const
+      {
+          return !pred(x);
+      }
+
+      __host__ __device__ __forceinline__ unary_negate() {}
+      __host__ __device__ __forceinline__ unary_negate(const unary_negate& other) : pred(other.pred) {}
+
+      Predicate pred;
+    };
+
+    template <typename Predicate> __host__ __device__ __forceinline__ unary_negate<Predicate> not1(const Predicate& pred)
+    {
+        return unary_negate<Predicate>(pred);
+    }
+
+    template <typename Predicate> struct binary_negate : binary_function<typename Predicate::first_argument_type, typename Predicate::second_argument_type, bool>
+    {
+        explicit __host__ __device__ __forceinline__ binary_negate(const Predicate& p) : pred(p) {}
+
+        __device__ __forceinline__ bool operator()(typename TypeTraits<typename Predicate::first_argument_type>::ParameterType x,
+                                                   typename TypeTraits<typename Predicate::second_argument_type>::ParameterType y) const
+        {
+            return !pred(x,y);
+        }
+
+        __host__ __device__ __forceinline__ binary_negate() {}
+        __host__ __device__ __forceinline__ binary_negate(const binary_negate& other) : pred(other.pred) {}
+
+        Predicate pred;
+    };
+
+    template <typename BinaryPredicate> __host__ __device__ __forceinline__ binary_negate<BinaryPredicate> not2(const BinaryPredicate& pred)
+    {
+        return binary_negate<BinaryPredicate>(pred);
+    }
+
+    template <typename Op> struct binder1st : unary_function<typename Op::second_argument_type, typename Op::result_type>
+    {
+        __host__ __device__ __forceinline__ binder1st(const Op& op_, const typename Op::first_argument_type& arg1_) : op(op_), arg1(arg1_) {}
+
+        __device__ __forceinline__ typename Op::result_type operator ()(typename TypeTraits<typename Op::second_argument_type>::ParameterType a) const
+        {
+            return op(arg1, a);
+        }
+
+        __host__ __device__ __forceinline__ binder1st() {}
+        __host__ __device__ __forceinline__ binder1st(const binder1st& other) : op(other.op), arg1(other.arg1) {}
+
+        Op op;
+        typename Op::first_argument_type arg1;
+    };
+
+    template <typename Op, typename T> __host__ __device__ __forceinline__ binder1st<Op> bind1st(const Op& op, const T& x)
+    {
+        return binder1st<Op>(op, typename Op::first_argument_type(x));
+    }
+
+    template <typename Op> struct binder2nd : unary_function<typename Op::first_argument_type, typename Op::result_type>
+    {
+        __host__ __device__ __forceinline__ binder2nd(const Op& op_, const typename Op::second_argument_type& arg2_) : op(op_), arg2(arg2_) {}
+
+        __forceinline__ __device__ typename Op::result_type operator ()(typename TypeTraits<typename Op::first_argument_type>::ParameterType a) const
+        {
+            return op(a, arg2);
+        }
+
+        __host__ __device__ __forceinline__ binder2nd() {}
+        __host__ __device__ __forceinline__ binder2nd(const binder2nd& other) : op(other.op), arg2(other.arg2) {}
+
+        Op op;
+        typename Op::second_argument_type arg2;
+    };
+
+    template <typename Op, typename T> __host__ __device__ __forceinline__ binder2nd<Op> bind2nd(const Op& op, const T& x)
+    {
+        return binder2nd<Op>(op, typename Op::second_argument_type(x));
+    }
+
+    // Functor Traits
+    template <typename F> struct IsUnaryFunction
+    {
+        typedef char Yes;
+        struct No {Yes a[2];};
+
+        template <typename T, typename D> static Yes check(unary_function<T, D>);
+        static No check(...);
+
+        static F makeF();
+
+        enum { value = (sizeof(check(makeF())) == sizeof(Yes)) };
+    };
+
+    template <typename F> struct IsBinaryFunction
+    {
+        typedef char Yes;
+        struct No {Yes a[2];};
+
+        template <typename T1, typename T2, typename D> static Yes check(binary_function<T1, T2, D>);
+        static No check(...);
+
+        static F makeF();
+
+        enum { value = (sizeof(check(makeF())) == sizeof(Yes)) };
+    };
+
+    namespace functional_detail
+    {
+        template <size_t src_elem_size, size_t dst_elem_size> struct UnOpShift { enum { shift = 1 }; };
+        template <size_t src_elem_size> struct UnOpShift<src_elem_size, 1> { enum { shift = 4 }; };
+        template <size_t src_elem_size> struct UnOpShift<src_elem_size, 2> { enum { shift = 2 }; };
+
+        template <typename T, typename D> struct DefaultUnaryShift
+        {
+            enum { shift = UnOpShift<sizeof(T), sizeof(D)>::shift };
+        };
+
+        template <size_t src_elem_size1, size_t src_elem_size2, size_t dst_elem_size> struct BinOpShift { enum { shift = 1 }; };
+        template <size_t src_elem_size1, size_t src_elem_size2> struct BinOpShift<src_elem_size1, src_elem_size2, 1> { enum { shift = 4 }; };
+        template <size_t src_elem_size1, size_t src_elem_size2> struct BinOpShift<src_elem_size1, src_elem_size2, 2> { enum { shift = 2 }; };
+
+        template <typename T1, typename T2, typename D> struct DefaultBinaryShift
+        {
+            enum { shift = BinOpShift<sizeof(T1), sizeof(T2), sizeof(D)>::shift };
+        };
+
+        template <typename Func, bool unary = IsUnaryFunction<Func>::value> struct ShiftDispatcher;
+        template <typename Func> struct ShiftDispatcher<Func, true>
+        {
+            enum { shift = DefaultUnaryShift<typename Func::argument_type, typename Func::result_type>::shift };
+        };
+        template <typename Func> struct ShiftDispatcher<Func, false>
+        {
+            enum { shift = DefaultBinaryShift<typename Func::first_argument_type, typename Func::second_argument_type, typename Func::result_type>::shift };
+        };
+    }
+
+    template <typename Func> struct DefaultTransformShift
+    {
+        enum { shift = functional_detail::ShiftDispatcher<Func>::shift };
+    };
+
+    template <typename Func> struct DefaultTransformFunctorTraits
+    {
+        enum { simple_block_dim_x = 16 };
+        enum { simple_block_dim_y = 16 };
+
+        enum { smart_block_dim_x = 16 };
+        enum { smart_block_dim_y = 16 };
+        enum { smart_shift = DefaultTransformShift<Func>::shift };
+    };
+
+    template <typename Func> struct TransformFunctorTraits : DefaultTransformFunctorTraits<Func> {};
+
+#define OPENCV_CUDA_TRANSFORM_FUNCTOR_TRAITS(type) \
+    template <> struct TransformFunctorTraits< type > : DefaultTransformFunctorTraits< type >
+}}} // namespace cv { namespace cuda { namespace cudev
+
+//! @endcond
+
+#endif // OPENCV_CUDA_FUNCTIONAL_HPP

model_zoo/official/lite/app/src/main/cpp/include/opencv2/core/cuda/limits.hpp

@@ -0,0 +1,128 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CUDA_LIMITS_HPP
+#define OPENCV_CUDA_LIMITS_HPP
+
+#include <limits.h>
+#include <float.h>
+#include "common.hpp"
+
+/** @file
+ * @deprecated Use @ref cudev instead.
+ */
+
+//! @cond IGNORED
+
+namespace cv { namespace cuda { namespace device
+{
+template <class T> struct numeric_limits;
+
+template <> struct numeric_limits<bool>
+{
+    __device__ __forceinline__ static bool min() { return false; }
+    __device__ __forceinline__ static bool max() { return true;  }
+    static const bool is_signed = false;
+};
+
+template <> struct numeric_limits<signed char>
+{
+    __device__ __forceinline__ static signed char min() { return SCHAR_MIN; }
+    __device__ __forceinline__ static signed char max() { return SCHAR_MAX; }
+    static const bool is_signed = true;
+};
+
+template <> struct numeric_limits<unsigned char>
+{
+    __device__ __forceinline__ static unsigned char min() { return 0; }
+    __device__ __forceinline__ static unsigned char max() { return UCHAR_MAX; }
+    static const bool is_signed = false;
+};
+
+template <> struct numeric_limits<short>
+{
+    __device__ __forceinline__ static short min() { return SHRT_MIN; }
+    __device__ __forceinline__ static short max() { return SHRT_MAX; }
+    static const bool is_signed = true;
+};
+
+template <> struct numeric_limits<unsigned short>
+{
+    __device__ __forceinline__ static unsigned short min() { return 0; }
+    __device__ __forceinline__ static unsigned short max() { return USHRT_MAX; }
+    static const bool is_signed = false;
+};
+
+template <> struct numeric_limits<int>
+{
+    __device__ __forceinline__ static int min() { return INT_MIN; }
+    __device__ __forceinline__ static int max() { return INT_MAX; }
+    static const bool is_signed = true;
+};
+
+template <> struct numeric_limits<unsigned int>
+{
+    __device__ __forceinline__ static unsigned int min() { return 0; }
+    __device__ __forceinline__ static unsigned int max() { return UINT_MAX; }
+    static const bool is_signed = false;
+};
+
+template <> struct numeric_limits<float>
+{
+    __device__ __forceinline__ static float min() { return FLT_MIN; }
+    __device__ __forceinline__ static float max() { return FLT_MAX; }
+    __device__ __forceinline__ static float epsilon() { return FLT_EPSILON; }
+    static const bool is_signed = true;
+};
+
+template <> struct numeric_limits<double>
+{
+    __device__ __forceinline__ static double min() { return DBL_MIN; }
+    __device__ __forceinline__ static double max() { return DBL_MAX; }
+    __device__ __forceinline__ static double epsilon() { return DBL_EPSILON; }
+    static const bool is_signed = true;
+};
+}}} // namespace cv { namespace cuda { namespace cudev {
+
+//! @endcond
+
+#endif // OPENCV_CUDA_LIMITS_HPP

model_zoo/official/lite/app/src/main/cpp/include/opencv2/core/cuda/reduce.hpp

@@ -0,0 +1,209 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CUDA_REDUCE_HPP
+#define OPENCV_CUDA_REDUCE_HPP
+
+#ifndef THRUST_DEBUG // eliminate -Wundef warning
+#define THRUST_DEBUG 0
+#endif
+
+#include <thrust/tuple.h>
+#include "detail/reduce.hpp"
+#include "detail/reduce_key_val.hpp"
+
+/** @file
+ * @deprecated Use @ref cudev instead.
+ */
+
+//! @cond IGNORED
+
+namespace cv { namespace cuda { namespace device
+{
+    template <int N, typename T, class Op>
+    __device__ __forceinline__ void reduce(volatile T* smem, T& val, unsigned int tid, const Op& op)
+    {
+        reduce_detail::Dispatcher<N>::reductor::template reduce<volatile T*, T&, const Op&>(smem, val, tid, op);
+    }
+    template <int N,
+              typename P0, typename P1, typename P2, typename P3, typename P4, typename P5, typename P6, typename P7, typename P8, typename P9,
+              typename R0, typename R1, typename R2, typename R3, typename R4, typename R5, typename R6, typename R7, typename R8, typename R9,
+              class Op0, class Op1, class Op2, class Op3, class Op4, class Op5, class Op6, class Op7, class Op8, class Op9>
+    __device__ __forceinline__ void reduce(const thrust::tuple<P0, P1, P2, P3, P4, P5, P6, P7, P8, P9>& smem,
+                                           const thrust::tuple<R0, R1, R2, R3, R4, R5, R6, R7, R8, R9>& val,
+                                           unsigned int tid,
+                                           const thrust::tuple<Op0, Op1, Op2, Op3, Op4, Op5, Op6, Op7, Op8, Op9>& op)
+    {
+        reduce_detail::Dispatcher<N>::reductor::template reduce<
+                const thrust::tuple<P0, P1, P2, P3, P4, P5, P6, P7, P8, P9>&,
+                const thrust::tuple<R0, R1, R2, R3, R4, R5, R6, R7, R8, R9>&,
+                const thrust::tuple<Op0, Op1, Op2, Op3, Op4, Op5, Op6, Op7, Op8, Op9>&>(smem, val, tid, op);
+    }
+
+    template <unsigned int N, typename K, typename V, class Cmp>
+    __device__ __forceinline__ void reduceKeyVal(volatile K* skeys, K& key, volatile V* svals, V& val, unsigned int tid, const Cmp& cmp)
+    {
+        reduce_key_val_detail::Dispatcher<N>::reductor::template reduce<volatile K*, K&, volatile V*, V&, const Cmp&>(skeys, key, svals, val, tid, cmp);
+    }
+    template <unsigned int N,
+              typename K,
+              typename VP0, typename VP1, typename VP2, typename VP3, typename VP4, typename VP5, typename VP6, typename VP7, typename VP8, typename VP9,
+              typename VR0, typename VR1, typename VR2, typename VR3, typename VR4, typename VR5, typename VR6, typename VR7, typename VR8, typename VR9,
+              class Cmp>
+    __device__ __forceinline__ void reduceKeyVal(volatile K* skeys, K& key,
+                                                 const thrust::tuple<VP0, VP1, VP2, VP3, VP4, VP5, VP6, VP7, VP8, VP9>& svals,
+                                                 const thrust::tuple<VR0, VR1, VR2, VR3, VR4, VR5, VR6, VR7, VR8, VR9>& val,
+                                                 unsigned int tid, const Cmp& cmp)
+    {
+        reduce_key_val_detail::Dispatcher<N>::reductor::template reduce<volatile K*, K&,
+                const thrust::tuple<VP0, VP1, VP2, VP3, VP4, VP5, VP6, VP7, VP8, VP9>&,
+                const thrust::tuple<VR0, VR1, VR2, VR3, VR4, VR5, VR6, VR7, VR8, VR9>&,
+                const Cmp&>(skeys, key, svals, val, tid, cmp);
+    }
+    template <unsigned int N,
+              typename KP0, typename KP1, typename KP2, typename KP3, typename KP4, typename KP5, typename KP6, typename KP7, typename KP8, typename KP9,
+              typename KR0, typename KR1, typename KR2, typename KR3, typename KR4, typename KR5, typename KR6, typename KR7, typename KR8, typename KR9,
+              typename VP0, typename VP1, typename VP2, typename VP3, typename VP4, typename VP5, typename VP6, typename VP7, typename VP8, typename VP9,
+              typename VR0, typename VR1, typename VR2, typename VR3, typename VR4, typename VR5, typename VR6, typename VR7, typename VR8, typename VR9,
+              class Cmp0, class Cmp1, class Cmp2, class Cmp3, class Cmp4, class Cmp5, class Cmp6, class Cmp7, class Cmp8, class Cmp9>
+    __device__ __forceinline__ void reduceKeyVal(const thrust::tuple<KP0, KP1, KP2, KP3, KP4, KP5, KP6, KP7, KP8, KP9>& skeys,
+                                                 const thrust::tuple<KR0, KR1, KR2, KR3, KR4, KR5, KR6, KR7, KR8, KR9>& key,
+                                                 const thrust::tuple<VP0, VP1, VP2, VP3, VP4, VP5, VP6, VP7, VP8, VP9>& svals,
+                                                 const thrust::tuple<VR0, VR1, VR2, VR3, VR4, VR5, VR6, VR7, VR8, VR9>& val,
+                                                 unsigned int tid,
+                                                 const thrust::tuple<Cmp0, Cmp1, Cmp2, Cmp3, Cmp4, Cmp5, Cmp6, Cmp7, Cmp8, Cmp9>& cmp)
+    {
+        reduce_key_val_detail::Dispatcher<N>::reductor::template reduce<
+                const thrust::tuple<KP0, KP1, KP2, KP3, KP4, KP5, KP6, KP7, KP8, KP9>&,
+                const thrust::tuple<KR0, KR1, KR2, KR3, KR4, KR5, KR6, KR7, KR8, KR9>&,
+                const thrust::tuple<VP0, VP1, VP2, VP3, VP4, VP5, VP6, VP7, VP8, VP9>&,
+                const thrust::tuple<VR0, VR1, VR2, VR3, VR4, VR5, VR6, VR7, VR8, VR9>&,
+                const thrust::tuple<Cmp0, Cmp1, Cmp2, Cmp3, Cmp4, Cmp5, Cmp6, Cmp7, Cmp8, Cmp9>&
+                >(skeys, key, svals, val, tid, cmp);
+    }
+
+    // smem_tuple
+
+    template <typename T0>
+    __device__ __forceinline__
+    thrust::tuple<volatile T0*>
+    smem_tuple(T0* t0)
+    {
+        return thrust::make_tuple((volatile T0*) t0);
+    }
+
+    template <typename T0, typename T1>
+    __device__ __forceinline__
+    thrust::tuple<volatile T0*, volatile T1*>
+    smem_tuple(T0* t0, T1* t1)
+    {
+        return thrust::make_tuple((volatile T0*) t0, (volatile T1*) t1);
+    }
+
+    template <typename T0, typename T1, typename T2>
+    __device__ __forceinline__
+    thrust::tuple<volatile T0*, volatile T1*, volatile T2*>
+    smem_tuple(T0* t0, T1* t1, T2* t2)
+    {
+        return thrust::make_tuple((volatile T0*) t0, (volatile T1*) t1, (volatile T2*) t2);
+    }
+
+    template <typename T0, typename T1, typename T2, typename T3>
+    __device__ __forceinline__
+    thrust::tuple<volatile T0*, volatile T1*, volatile T2*, volatile T3*>
+    smem_tuple(T0* t0, T1* t1, T2* t2, T3* t3)
+    {
+        return thrust::make_tuple((volatile T0*) t0, (volatile T1*) t1, (volatile T2*) t2, (volatile T3*) t3);
+    }
+
+    template <typename T0, typename T1, typename T2, typename T3, typename T4>
+    __device__ __forceinline__
+    thrust::tuple<volatile T0*, volatile T1*, volatile T2*, volatile T3*, volatile T4*>
+    smem_tuple(T0* t0, T1* t1, T2* t2, T3* t3, T4* t4)
+    {
+        return thrust::make_tuple((volatile T0*) t0, (volatile T1*) t1, (volatile T2*) t2, (volatile T3*) t3, (volatile T4*) t4);
+    }
+
+    template <typename T0, typename T1, typename T2, typename T3, typename T4, typename T5>
+    __device__ __forceinline__
+    thrust::tuple<volatile T0*, volatile T1*, volatile T2*, volatile T3*, volatile T4*, volatile T5*>
+    smem_tuple(T0* t0, T1* t1, T2* t2, T3* t3, T4* t4, T5* t5)
+    {
+        return thrust::make_tuple((volatile T0*) t0, (volatile T1*) t1, (volatile T2*) t2, (volatile T3*) t3, (volatile T4*) t4, (volatile T5*) t5);
+    }
+
+    template <typename T0, typename T1, typename T2, typename T3, typename T4, typename T5, typename T6>
+    __device__ __forceinline__
+    thrust::tuple<volatile T0*, volatile T1*, volatile T2*, volatile T3*, volatile T4*, volatile T5*, volatile T6*>
+    smem_tuple(T0* t0, T1* t1, T2* t2, T3* t3, T4* t4, T5* t5, T6* t6)
+    {
+        return thrust::make_tuple((volatile T0*) t0, (volatile T1*) t1, (volatile T2*) t2, (volatile T3*) t3, (volatile T4*) t4, (volatile T5*) t5, (volatile T6*) t6);
+    }
+
+    template <typename T0, typename T1, typename T2, typename T3, typename T4, typename T5, typename T6, typename T7>
+    __device__ __forceinline__
+    thrust::tuple<volatile T0*, volatile T1*, volatile T2*, volatile T3*, volatile T4*, volatile T5*, volatile T6*, volatile T7*>
+    smem_tuple(T0* t0, T1* t1, T2* t2, T3* t3, T4* t4, T5* t5, T6* t6, T7* t7)
+    {
+        return thrust::make_tuple((volatile T0*) t0, (volatile T1*) t1, (volatile T2*) t2, (volatile T3*) t3, (volatile T4*) t4, (volatile T5*) t5, (volatile T6*) t6, (volatile T7*) t7);
+    }
+
+    template <typename T0, typename T1, typename T2, typename T3, typename T4, typename T5, typename T6, typename T7, typename T8>
+    __device__ __forceinline__
+    thrust::tuple<volatile T0*, volatile T1*, volatile T2*, volatile T3*, volatile T4*, volatile T5*, volatile T6*, volatile T7*, volatile T8*>
+    smem_tuple(T0* t0, T1* t1, T2* t2, T3* t3, T4* t4, T5* t5, T6* t6, T7* t7, T8* t8)
+    {
+        return thrust::make_tuple((volatile T0*) t0, (volatile T1*) t1, (volatile T2*) t2, (volatile T3*) t3, (volatile T4*) t4, (volatile T5*) t5, (volatile T6*) t6, (volatile T7*) t7, (volatile T8*) t8);
+    }
+
+    template <typename T0, typename T1, typename T2, typename T3, typename T4, typename T5, typename T6, typename T7, typename T8, typename T9>
+    __device__ __forceinline__
+    thrust::tuple<volatile T0*, volatile T1*, volatile T2*, volatile T3*, volatile T4*, volatile T5*, volatile T6*, volatile T7*, volatile T8*, volatile T9*>
+    smem_tuple(T0* t0, T1* t1, T2* t2, T3* t3, T4* t4, T5* t5, T6* t6, T7* t7, T8* t8, T9* t9)
+    {
+        return thrust::make_tuple((volatile T0*) t0, (volatile T1*) t1, (volatile T2*) t2, (volatile T3*) t3, (volatile T4*) t4, (volatile T5*) t5, (volatile T6*) t6, (volatile T7*) t7, (volatile T8*) t8, (volatile T9*) t9);
+    }
+}}}
+
+//! @endcond
+
+#endif // OPENCV_CUDA_REDUCE_HPP

model_zoo/official/lite/app/src/main/cpp/include/opencv2/core/cuda/saturate_cast.hpp

@@ -0,0 +1,292 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CUDA_SATURATE_CAST_HPP
+#define OPENCV_CUDA_SATURATE_CAST_HPP
+
+#include "common.hpp"
+
+/** @file
+ * @deprecated Use @ref cudev instead.
+ */
+
+//! @cond IGNORED
+
+namespace cv { namespace cuda { namespace device
+{
+    template<typename _Tp> __device__ __forceinline__ _Tp saturate_cast(uchar v) { return _Tp(v); }
+    template<typename _Tp> __device__ __forceinline__ _Tp saturate_cast(schar v) { return _Tp(v); }
+    template<typename _Tp> __device__ __forceinline__ _Tp saturate_cast(ushort v) { return _Tp(v); }
+    template<typename _Tp> __device__ __forceinline__ _Tp saturate_cast(short v) { return _Tp(v); }
+    template<typename _Tp> __device__ __forceinline__ _Tp saturate_cast(uint v) { return _Tp(v); }
+    template<typename _Tp> __device__ __forceinline__ _Tp saturate_cast(int v) { return _Tp(v); }
+    template<typename _Tp> __device__ __forceinline__ _Tp saturate_cast(float v) { return _Tp(v); }
+    template<typename _Tp> __device__ __forceinline__ _Tp saturate_cast(double v) { return _Tp(v); }
+
+    template<> __device__ __forceinline__ uchar saturate_cast<uchar>(schar v)
+    {
+        uint res = 0;
+        int vi = v;
+        asm("cvt.sat.u8.s8 %0, %1;" : "=r"(res) : "r"(vi));
+        return res;
+    }
+    template<> __device__ __forceinline__ uchar saturate_cast<uchar>(short v)
+    {
+        uint res = 0;
+        asm("cvt.sat.u8.s16 %0, %1;" : "=r"(res) : "h"(v));
+        return res;
+    }
+    template<> __device__ __forceinline__ uchar saturate_cast<uchar>(ushort v)
+    {
+        uint res = 0;
+        asm("cvt.sat.u8.u16 %0, %1;" : "=r"(res) : "h"(v));
+        return res;
+    }
+    template<> __device__ __forceinline__ uchar saturate_cast<uchar>(int v)
+    {
+        uint res = 0;
+        asm("cvt.sat.u8.s32 %0, %1;" : "=r"(res) : "r"(v));
+        return res;
+    }
+    template<> __device__ __forceinline__ uchar saturate_cast<uchar>(uint v)
+    {
+        uint res = 0;
+        asm("cvt.sat.u8.u32 %0, %1;" : "=r"(res) : "r"(v));
+        return res;
+    }
+    template<> __device__ __forceinline__ uchar saturate_cast<uchar>(float v)
+    {
+        uint res = 0;
+        asm("cvt.rni.sat.u8.f32 %0, %1;" : "=r"(res) : "f"(v));
+        return res;
+    }
+    template<> __device__ __forceinline__ uchar saturate_cast<uchar>(double v)
+    {
+    #if defined __CUDA_ARCH__ && __CUDA_ARCH__ >= 130
+        uint res = 0;
+        asm("cvt.rni.sat.u8.f64 %0, %1;" : "=r"(res) : "d"(v));
+        return res;
+    #else
+        return saturate_cast<uchar>((float)v);
+    #endif
+    }
+
+    template<> __device__ __forceinline__ schar saturate_cast<schar>(uchar v)
+    {
+        uint res = 0;
+        uint vi = v;
+        asm("cvt.sat.s8.u8 %0, %1;" : "=r"(res) : "r"(vi));
+        return res;
+    }
+    template<> __device__ __forceinline__ schar saturate_cast<schar>(short v)
+    {
+        uint res = 0;
+        asm("cvt.sat.s8.s16 %0, %1;" : "=r"(res) : "h"(v));
+        return res;
+    }
+    template<> __device__ __forceinline__ schar saturate_cast<schar>(ushort v)
+    {
+        uint res = 0;
+        asm("cvt.sat.s8.u16 %0, %1;" : "=r"(res) : "h"(v));
+        return res;
+    }
+    template<> __device__ __forceinline__ schar saturate_cast<schar>(int v)
+    {
+        uint res = 0;
+        asm("cvt.sat.s8.s32 %0, %1;" : "=r"(res) : "r"(v));
+        return res;
+    }
+    template<> __device__ __forceinline__ schar saturate_cast<schar>(uint v)
+    {
+        uint res = 0;
+        asm("cvt.sat.s8.u32 %0, %1;" : "=r"(res) : "r"(v));
+        return res;
+    }
+    template<> __device__ __forceinline__ schar saturate_cast<schar>(float v)
+    {
+        uint res = 0;
+        asm("cvt.rni.sat.s8.f32 %0, %1;" : "=r"(res) : "f"(v));
+        return res;
+    }
+    template<> __device__ __forceinline__ schar saturate_cast<schar>(double v)
+    {
+    #if defined __CUDA_ARCH__ && __CUDA_ARCH__ >= 130
+        uint res = 0;
+        asm("cvt.rni.sat.s8.f64 %0, %1;" : "=r"(res) : "d"(v));
+        return res;
+    #else
+        return saturate_cast<schar>((float)v);
+    #endif
+    }
+
+    template<> __device__ __forceinline__ ushort saturate_cast<ushort>(schar v)
+    {
+        ushort res = 0;
+        int vi = v;
+        asm("cvt.sat.u16.s8 %0, %1;" : "=h"(res) : "r"(vi));
+        return res;
+    }
+    template<> __device__ __forceinline__ ushort saturate_cast<ushort>(short v)
+    {
+        ushort res = 0;
+        asm("cvt.sat.u16.s16 %0, %1;" : "=h"(res) : "h"(v));
+        return res;
+    }
+    template<> __device__ __forceinline__ ushort saturate_cast<ushort>(int v)
+    {
+        ushort res = 0;
+        asm("cvt.sat.u16.s32 %0, %1;" : "=h"(res) : "r"(v));
+        return res;
+    }
+    template<> __device__ __forceinline__ ushort saturate_cast<ushort>(uint v)
+    {
+        ushort res = 0;
+        asm("cvt.sat.u16.u32 %0, %1;" : "=h"(res) : "r"(v));
+        return res;
+    }
+    template<> __device__ __forceinline__ ushort saturate_cast<ushort>(float v)
+    {
+        ushort res = 0;
+        asm("cvt.rni.sat.u16.f32 %0, %1;" : "=h"(res) : "f"(v));
+        return res;
+    }
+    template<> __device__ __forceinline__ ushort saturate_cast<ushort>(double v)
+    {
+    #if defined __CUDA_ARCH__ && __CUDA_ARCH__ >= 130
+        ushort res = 0;
+        asm("cvt.rni.sat.u16.f64 %0, %1;" : "=h"(res) : "d"(v));
+        return res;
+    #else
+        return saturate_cast<ushort>((float)v);
+    #endif
+    }
+
+    template<> __device__ __forceinline__ short saturate_cast<short>(ushort v)
+    {
+        short res = 0;
+        asm("cvt.sat.s16.u16 %0, %1;" : "=h"(res) : "h"(v));
+        return res;
+    }
+    template<> __device__ __forceinline__ short saturate_cast<short>(int v)
+    {
+        short res = 0;
+        asm("cvt.sat.s16.s32 %0, %1;" : "=h"(res) : "r"(v));
+        return res;
+    }
+    template<> __device__ __forceinline__ short saturate_cast<short>(uint v)
+    {
+        short res = 0;
+        asm("cvt.sat.s16.u32 %0, %1;" : "=h"(res) : "r"(v));
+        return res;
+    }
+    template<> __device__ __forceinline__ short saturate_cast<short>(float v)
+    {
+        short res = 0;
+        asm("cvt.rni.sat.s16.f32 %0, %1;" : "=h"(res) : "f"(v));
+        return res;
+    }
+    template<> __device__ __forceinline__ short saturate_cast<short>(double v)
+    {
+    #if defined __CUDA_ARCH__ && __CUDA_ARCH__ >= 130
+        short res = 0;
+        asm("cvt.rni.sat.s16.f64 %0, %1;" : "=h"(res) : "d"(v));
+        return res;
+    #else
+        return saturate_cast<short>((float)v);
+    #endif
+    }
+
+    template<> __device__ __forceinline__ int saturate_cast<int>(uint v)
+    {
+        int res = 0;
+        asm("cvt.sat.s32.u32 %0, %1;" : "=r"(res) : "r"(v));
+        return res;
+    }
+    template<> __device__ __forceinline__ int saturate_cast<int>(float v)
+    {
+        return __float2int_rn(v);
+    }
+    template<> __device__ __forceinline__ int saturate_cast<int>(double v)
+    {
+    #if defined __CUDA_ARCH__ && __CUDA_ARCH__ >= 130
+        return __double2int_rn(v);
+    #else
+        return saturate_cast<int>((float)v);
+    #endif
+    }
+
+    template<> __device__ __forceinline__ uint saturate_cast<uint>(schar v)
+    {
+        uint res = 0;
+        int vi = v;
+        asm("cvt.sat.u32.s8 %0, %1;" : "=r"(res) : "r"(vi));
+        return res;
+    }
+    template<> __device__ __forceinline__ uint saturate_cast<uint>(short v)
+    {
+        uint res = 0;
+        asm("cvt.sat.u32.s16 %0, %1;" : "=r"(res) : "h"(v));
+        return res;
+    }
+    template<> __device__ __forceinline__ uint saturate_cast<uint>(int v)
+    {
+        uint res = 0;
+        asm("cvt.sat.u32.s32 %0, %1;" : "=r"(res) : "r"(v));
+        return res;
+    }
+    template<> __device__ __forceinline__ uint saturate_cast<uint>(float v)
+    {
+        return __float2uint_rn(v);
+    }
+    template<> __device__ __forceinline__ uint saturate_cast<uint>(double v)
+    {
+    #if defined __CUDA_ARCH__ && __CUDA_ARCH__ >= 130
+        return __double2uint_rn(v);
+    #else
+        return saturate_cast<uint>((float)v);
+    #endif
+    }
+}}}
+
+//! @endcond
+
+#endif /* OPENCV_CUDA_SATURATE_CAST_HPP */

model_zoo/official/lite/app/src/main/cpp/include/opencv2/core/cuda/scan.hpp

@@ -0,0 +1,258 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CUDA_SCAN_HPP
+#define OPENCV_CUDA_SCAN_HPP
+
+#include "opencv2/core/cuda/common.hpp"
+#include "opencv2/core/cuda/utility.hpp"
+#include "opencv2/core/cuda/warp.hpp"
+#include "opencv2/core/cuda/warp_shuffle.hpp"
+
+/** @file
+ * @deprecated Use @ref cudev instead.
+ */
+
+//! @cond IGNORED
+
+namespace cv { namespace cuda { namespace device
+{
+    enum ScanKind { EXCLUSIVE = 0,  INCLUSIVE = 1 };
+
+    template <ScanKind Kind, typename T, typename F> struct WarpScan
+    {
+        __device__ __forceinline__ WarpScan() {}
+        __device__ __forceinline__ WarpScan(const WarpScan& other) { CV_UNUSED(other); }
+
+        __device__ __forceinline__ T operator()( volatile T *ptr , const unsigned int idx)
+        {
+            const unsigned int lane = idx & 31;
+            F op;
+
+            if ( lane >=  1) ptr [idx ] = op(ptr [idx -  1], ptr [idx]);
+            if ( lane >=  2) ptr [idx ] = op(ptr [idx -  2], ptr [idx]);
+            if ( lane >=  4) ptr [idx ] = op(ptr [idx -  4], ptr [idx]);
+            if ( lane >=  8) ptr [idx ] = op(ptr [idx -  8], ptr [idx]);
+            if ( lane >= 16) ptr [idx ] = op(ptr [idx - 16], ptr [idx]);
+
+            if( Kind == INCLUSIVE )
+                return ptr [idx];
+            else
+                return (lane > 0) ? ptr [idx - 1] : 0;
+        }
+
+        __device__ __forceinline__ unsigned int index(const unsigned int tid)
+        {
+            return tid;
+        }
+
+        __device__ __forceinline__ void init(volatile T *ptr){}
+
+        static const int warp_offset      = 0;
+
+        typedef WarpScan<INCLUSIVE, T, F>  merge;
+    };
+
+    template <ScanKind Kind , typename T, typename F> struct WarpScanNoComp
+    {
+        __device__ __forceinline__ WarpScanNoComp() {}
+        __device__ __forceinline__ WarpScanNoComp(const WarpScanNoComp& other) { CV_UNUSED(other); }
+
+        __device__ __forceinline__ T operator()( volatile T *ptr , const unsigned int idx)
+        {
+            const unsigned int lane = threadIdx.x & 31;
+            F op;
+
+            ptr [idx ] = op(ptr [idx -  1], ptr [idx]);
+            ptr [idx ] = op(ptr [idx -  2], ptr [idx]);
+            ptr [idx ] = op(ptr [idx -  4], ptr [idx]);
+            ptr [idx ] = op(ptr [idx -  8], ptr [idx]);
+            ptr [idx ] = op(ptr [idx - 16], ptr [idx]);
+
+            if( Kind == INCLUSIVE )
+                return ptr [idx];
+            else
+                return (lane > 0) ? ptr [idx - 1] : 0;
+        }
+
+        __device__ __forceinline__ unsigned int index(const unsigned int tid)
+        {
+            return (tid >> warp_log) * warp_smem_stride + 16 + (tid & warp_mask);
+        }
+
+        __device__ __forceinline__ void init(volatile T *ptr)
+        {
+            ptr[threadIdx.x] = 0;
+        }
+
+        static const int warp_smem_stride = 32 + 16 + 1;
+        static const int warp_offset      = 16;
+        static const int warp_log         = 5;
+        static const int warp_mask        = 31;
+
+        typedef WarpScanNoComp<INCLUSIVE, T, F> merge;
+    };
+
+    template <ScanKind Kind , typename T, typename Sc, typename F> struct BlockScan
+    {
+        __device__ __forceinline__ BlockScan() {}
+        __device__ __forceinline__ BlockScan(const BlockScan& other) { CV_UNUSED(other); }
+
+        __device__ __forceinline__ T operator()(volatile T *ptr)
+        {
+            const unsigned int tid  = threadIdx.x;
+            const unsigned int lane = tid & warp_mask;
+            const unsigned int warp = tid >> warp_log;
+
+            Sc scan;
+            typename Sc::merge merge_scan;
+            const unsigned int idx = scan.index(tid);
+
+            T val = scan(ptr, idx);
+            __syncthreads ();
+
+            if( warp == 0)
+                scan.init(ptr);
+            __syncthreads ();
+
+            if( lane == 31 )
+                ptr [scan.warp_offset + warp ] = (Kind == INCLUSIVE) ? val : ptr [idx];
+            __syncthreads ();
+
+            if( warp == 0 )
+                merge_scan(ptr, idx);
+            __syncthreads();
+
+            if ( warp > 0)
+                val = ptr [scan.warp_offset + warp - 1] + val;
+            __syncthreads ();
+
+            ptr[idx] = val;
+            __syncthreads ();
+
+            return val ;
+        }
+
+        static const int warp_log  = 5;
+        static const int warp_mask = 31;
+    };
+
+    template <typename T>
+    __device__ T warpScanInclusive(T idata, volatile T* s_Data, unsigned int tid)
+    {
+    #if __CUDA_ARCH__ >= 300
+        const unsigned int laneId = cv::cuda::device::Warp::laneId();
+
+        // scan on shuffl functions
+        #pragma unroll
+        for (int i = 1; i <= (OPENCV_CUDA_WARP_SIZE / 2); i *= 2)
+        {
+            const T n = cv::cuda::device::shfl_up(idata, i);
+            if (laneId >= i)
+                  idata += n;
+        }
+
+        return idata;
+    #else
+        unsigned int pos = 2 * tid - (tid & (OPENCV_CUDA_WARP_SIZE - 1));
+        s_Data[pos] = 0;
+        pos += OPENCV_CUDA_WARP_SIZE;
+        s_Data[pos] = idata;
+
+        s_Data[pos] += s_Data[pos - 1];
+        s_Data[pos] += s_Data[pos - 2];
+        s_Data[pos] += s_Data[pos - 4];
+        s_Data[pos] += s_Data[pos - 8];
+        s_Data[pos] += s_Data[pos - 16];
+
+        return s_Data[pos];
+    #endif
+    }
+
+    template <typename T>
+    __device__ __forceinline__ T warpScanExclusive(T idata, volatile T* s_Data, unsigned int tid)
+    {
+        return warpScanInclusive(idata, s_Data, tid) - idata;
+    }
+
+    template <int tiNumScanThreads, typename T>
+    __device__ T blockScanInclusive(T idata, volatile T* s_Data, unsigned int tid)
+    {
+        if (tiNumScanThreads > OPENCV_CUDA_WARP_SIZE)
+        {
+            //Bottom-level inclusive warp scan
+            T warpResult = warpScanInclusive(idata, s_Data, tid);
+
+            //Save top elements of each warp for exclusive warp scan
+            //sync to wait for warp scans to complete (because s_Data is being overwritten)
+            __syncthreads();
+            if ((tid & (OPENCV_CUDA_WARP_SIZE - 1)) == (OPENCV_CUDA_WARP_SIZE - 1))
+            {
+                s_Data[tid >> OPENCV_CUDA_LOG_WARP_SIZE] = warpResult;
+            }
+
+            //wait for warp scans to complete
+            __syncthreads();
+
+            if (tid < (tiNumScanThreads / OPENCV_CUDA_WARP_SIZE) )
+            {
+                //grab top warp elements
+                T val = s_Data[tid];
+                //calculate exclusive scan and write back to shared memory
+                s_Data[tid] = warpScanExclusive(val, s_Data, tid);
+            }
+
+            //return updated warp scans with exclusive scan results
+            __syncthreads();
+
+            return warpResult + s_Data[tid >> OPENCV_CUDA_LOG_WARP_SIZE];
+        }
+        else
+        {
+            return warpScanInclusive(idata, s_Data, tid);
+        }
+    }
+}}}
+
+//! @endcond
+
+#endif // OPENCV_CUDA_SCAN_HPP

model_zoo/official/lite/app/src/main/cpp/include/opencv2/core/cuda/simd_functions.hpp

@@ -0,0 +1,869 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+/*
+ * Copyright (c) 2013 NVIDIA Corporation. All rights reserved.
+ *
+ * Redistribution and use in source and binary forms, with or without
+ * modification, are permitted provided that the following conditions are met:
+ *
+ *   Redistributions of source code must retain the above copyright notice,
+ *   this list of conditions and the following disclaimer.
+ *
+ *   Redistributions in binary form must reproduce the above copyright notice,
+ *   this list of conditions and the following disclaimer in the documentation
+ *   and/or other materials provided with the distribution.
+ *
+ *   Neither the name of NVIDIA Corporation nor the names of its contributors
+ *   may be used to endorse or promote products derived from this software
+ *   without specific prior written permission.
+ *
+ * THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND CONTRIBUTORS "AS IS"
+ * AND ANY EXPRESS OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE
+ * IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE
+ * ARE DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT HOLDER OR CONTRIBUTORS BE
+ * LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR
+ * CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF
+ * SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS
+ * INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN
+ * CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE)
+ * ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE
+ * POSSIBILITY OF SUCH DAMAGE.
+ */
+
+#ifndef OPENCV_CUDA_SIMD_FUNCTIONS_HPP
+#define OPENCV_CUDA_SIMD_FUNCTIONS_HPP
+
+#include "common.hpp"
+
+/** @file
+ * @deprecated Use @ref cudev instead.
+ */
+
+//! @cond IGNORED
+
+namespace cv { namespace cuda { namespace device
+{
+    // 2
+
+    static __device__ __forceinline__ unsigned int vadd2(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vadd2.u32.u32.u32.sat %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #elif __CUDA_ARCH__ >= 200
+        asm("vadd.u32.u32.u32.sat %0.h0, %1.h0, %2.h0, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+        asm("vadd.u32.u32.u32.sat %0.h1, %1.h1, %2.h1, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        unsigned int s;
+        s = a ^ b;          // sum bits
+        r = a + b;          // actual sum
+        s = s ^ r;          // determine carry-ins for each bit position
+        s = s & 0x00010000; // carry-in to high word (= carry-out from low word)
+        r = r - s;          // subtract out carry-out from low word
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vsub2(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vsub2.u32.u32.u32.sat %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #elif __CUDA_ARCH__ >= 200
+        asm("vsub.u32.u32.u32.sat %0.h0, %1.h0, %2.h0, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+        asm("vsub.u32.u32.u32.sat %0.h1, %1.h1, %2.h1, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        unsigned int s;
+        s = a ^ b;          // sum bits
+        r = a - b;          // actual sum
+        s = s ^ r;          // determine carry-ins for each bit position
+        s = s & 0x00010000; // borrow to high word
+        r = r + s;          // compensate for borrow from low word
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vabsdiff2(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vabsdiff2.u32.u32.u32.sat %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #elif __CUDA_ARCH__ >= 200
+        asm("vabsdiff.u32.u32.u32.sat %0.h0, %1.h0, %2.h0, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+        asm("vabsdiff.u32.u32.u32.sat %0.h1, %1.h1, %2.h1, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        unsigned int s, t, u, v;
+        s = a & 0x0000ffff; // extract low halfword
+        r = b & 0x0000ffff; // extract low halfword
+        u = ::max(r, s);    // maximum of low halfwords
+        v = ::min(r, s);    // minimum of low halfwords
+        s = a & 0xffff0000; // extract high halfword
+        r = b & 0xffff0000; // extract high halfword
+        t = ::max(r, s);    // maximum of high halfwords
+        s = ::min(r, s);    // minimum of high halfwords
+        r = u | t;          // maximum of both halfwords
+        s = v | s;          // minimum of both halfwords
+        r = r - s;          // |a - b| = max(a,b) - min(a,b);
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vavg2(unsigned int a, unsigned int b)
+    {
+        unsigned int r, s;
+
+        // HAKMEM #23: a + b = 2 * (a & b) + (a ^ b) ==>
+        // (a + b) / 2 = (a & b) + ((a ^ b) >> 1)
+        s = a ^ b;
+        r = a & b;
+        s = s & 0xfffefffe; // ensure shift doesn't cross halfword boundaries
+        s = s >> 1;
+        s = r + s;
+
+        return s;
+    }
+
+    static __device__ __forceinline__ unsigned int vavrg2(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vavrg2.u32.u32.u32 %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        // HAKMEM #23: a + b = 2 * (a | b) - (a ^ b) ==>
+        // (a + b + 1) / 2 = (a | b) - ((a ^ b) >> 1)
+        unsigned int s;
+        s = a ^ b;
+        r = a | b;
+        s = s & 0xfffefffe; // ensure shift doesn't cross half-word boundaries
+        s = s >> 1;
+        r = r - s;
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vseteq2(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vset2.u32.u32.eq %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        // inspired by Alan Mycroft's null-byte detection algorithm:
+        // null_byte(x) = ((x - 0x01010101) & (~x & 0x80808080))
+        unsigned int c;
+        r = a ^ b;          // 0x0000 if a == b
+        c = r | 0x80008000; // set msbs, to catch carry out
+        r = r ^ c;          // extract msbs, msb = 1 if r < 0x8000
+        c = c - 0x00010001; // msb = 0, if r was 0x0000 or 0x8000
+        c = r & ~c;         // msb = 1, if r was 0x0000
+        r = c >> 15;        // convert to bool
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vcmpeq2(unsigned int a, unsigned int b)
+    {
+        unsigned int r, c;
+
+    #if __CUDA_ARCH__ >= 300
+        r = vseteq2(a, b);
+        c = r << 16;        // convert bool
+        r = c - r;          //  into mask
+    #else
+        // inspired by Alan Mycroft's null-byte detection algorithm:
+        // null_byte(x) = ((x - 0x01010101) & (~x & 0x80808080))
+        r = a ^ b;          // 0x0000 if a == b
+        c = r | 0x80008000; // set msbs, to catch carry out
+        r = r ^ c;          // extract msbs, msb = 1 if r < 0x8000
+        c = c - 0x00010001; // msb = 0, if r was 0x0000 or 0x8000
+        c = r & ~c;         // msb = 1, if r was 0x0000
+        r = c >> 15;        // convert
+        r = c - r;          //  msbs to
+        r = c | r;          //   mask
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vsetge2(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vset2.u32.u32.ge %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        unsigned int c;
+        asm("not.b32 %0, %0;" : "+r"(b));
+        c = vavrg2(a, b);   // (a + ~b + 1) / 2 = (a - b) / 2
+        c = c & 0x80008000; // msb = carry-outs
+        r = c >> 15;        // convert to bool
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vcmpge2(unsigned int a, unsigned int b)
+    {
+        unsigned int r, c;
+
+    #if __CUDA_ARCH__ >= 300
+        r = vsetge2(a, b);
+        c = r << 16;        // convert bool
+        r = c - r;          //  into mask
+    #else
+        asm("not.b32 %0, %0;" : "+r"(b));
+        c = vavrg2(a, b);   // (a + ~b + 1) / 2 = (a - b) / 2
+        c = c & 0x80008000; // msb = carry-outs
+        r = c >> 15;        // convert
+        r = c - r;          //  msbs to
+        r = c | r;          //   mask
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vsetgt2(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vset2.u32.u32.gt %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        unsigned int c;
+        asm("not.b32 %0, %0;" : "+r"(b));
+        c = vavg2(a, b);    // (a + ~b) / 2 = (a - b) / 2 [rounded down]
+        c = c & 0x80008000; // msbs = carry-outs
+        r = c >> 15;        // convert to bool
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vcmpgt2(unsigned int a, unsigned int b)
+    {
+        unsigned int r, c;
+
+    #if __CUDA_ARCH__ >= 300
+        r = vsetgt2(a, b);
+        c = r << 16;        // convert bool
+        r = c - r;          //  into mask
+    #else
+        asm("not.b32 %0, %0;" : "+r"(b));
+        c = vavg2(a, b);    // (a + ~b) / 2 = (a - b) / 2 [rounded down]
+        c = c & 0x80008000; // msbs = carry-outs
+        r = c >> 15;        // convert
+        r = c - r;          //  msbs to
+        r = c | r;          //   mask
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vsetle2(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vset2.u32.u32.le %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        unsigned int c;
+        asm("not.b32 %0, %0;" : "+r"(a));
+        c = vavrg2(a, b);   // (b + ~a + 1) / 2 = (b - a) / 2
+        c = c & 0x80008000; // msb = carry-outs
+        r = c >> 15;        // convert to bool
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vcmple2(unsigned int a, unsigned int b)
+    {
+        unsigned int r, c;
+
+    #if __CUDA_ARCH__ >= 300
+        r = vsetle2(a, b);
+        c = r << 16;        // convert bool
+        r = c - r;          //  into mask
+    #else
+        asm("not.b32 %0, %0;" : "+r"(a));
+        c = vavrg2(a, b);   // (b + ~a + 1) / 2 = (b - a) / 2
+        c = c & 0x80008000; // msb = carry-outs
+        r = c >> 15;        // convert
+        r = c - r;          //  msbs to
+        r = c | r;          //   mask
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vsetlt2(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vset2.u32.u32.lt %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        unsigned int c;
+        asm("not.b32 %0, %0;" : "+r"(a));
+        c = vavg2(a, b);    // (b + ~a) / 2 = (b - a) / 2 [rounded down]
+        c = c & 0x80008000; // msb = carry-outs
+        r = c >> 15;        // convert to bool
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vcmplt2(unsigned int a, unsigned int b)
+    {
+        unsigned int r, c;
+
+    #if __CUDA_ARCH__ >= 300
+        r = vsetlt2(a, b);
+        c = r << 16;        // convert bool
+        r = c - r;          //  into mask
+    #else
+        asm("not.b32 %0, %0;" : "+r"(a));
+        c = vavg2(a, b);    // (b + ~a) / 2 = (b - a) / 2 [rounded down]
+        c = c & 0x80008000; // msb = carry-outs
+        r = c >> 15;        // convert
+        r = c - r;          //  msbs to
+        r = c | r;          //   mask
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vsetne2(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm ("vset2.u32.u32.ne %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        // inspired by Alan Mycroft's null-byte detection algorithm:
+        // null_byte(x) = ((x - 0x01010101) & (~x & 0x80808080))
+        unsigned int c;
+        r = a ^ b;          // 0x0000 if a == b
+        c = r | 0x80008000; // set msbs, to catch carry out
+        c = c - 0x00010001; // msb = 0, if r was 0x0000 or 0x8000
+        c = r | c;          // msb = 1, if r was not 0x0000
+        c = c & 0x80008000; // extract msbs
+        r = c >> 15;        // convert to bool
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vcmpne2(unsigned int a, unsigned int b)
+    {
+        unsigned int r, c;
+
+    #if __CUDA_ARCH__ >= 300
+        r = vsetne2(a, b);
+        c = r << 16;        // convert bool
+        r = c - r;          //  into mask
+    #else
+        // inspired by Alan Mycroft's null-byte detection algorithm:
+        // null_byte(x) = ((x - 0x01010101) & (~x & 0x80808080))
+        r = a ^ b;          // 0x0000 if a == b
+        c = r | 0x80008000; // set msbs, to catch carry out
+        c = c - 0x00010001; // msb = 0, if r was 0x0000 or 0x8000
+        c = r | c;          // msb = 1, if r was not 0x0000
+        c = c & 0x80008000; // extract msbs
+        r = c >> 15;        // convert
+        r = c - r;          //  msbs to
+        r = c | r;          //   mask
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vmax2(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vmax2.u32.u32.u32 %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #elif __CUDA_ARCH__ >= 200
+        asm("vmax.u32.u32.u32 %0.h0, %1.h0, %2.h0, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+        asm("vmax.u32.u32.u32 %0.h1, %1.h1, %2.h1, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        unsigned int s, t, u;
+        r = a & 0x0000ffff; // extract low halfword
+        s = b & 0x0000ffff; // extract low halfword
+        t = ::max(r, s);    // maximum of low halfwords
+        r = a & 0xffff0000; // extract high halfword
+        s = b & 0xffff0000; // extract high halfword
+        u = ::max(r, s);    // maximum of high halfwords
+        r = t | u;          // combine halfword maximums
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vmin2(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vmin2.u32.u32.u32 %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #elif __CUDA_ARCH__ >= 200
+        asm("vmin.u32.u32.u32 %0.h0, %1.h0, %2.h0, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+        asm("vmin.u32.u32.u32 %0.h1, %1.h1, %2.h1, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        unsigned int s, t, u;
+        r = a & 0x0000ffff; // extract low halfword
+        s = b & 0x0000ffff; // extract low halfword
+        t = ::min(r, s);    // minimum of low halfwords
+        r = a & 0xffff0000; // extract high halfword
+        s = b & 0xffff0000; // extract high halfword
+        u = ::min(r, s);    // minimum of high halfwords
+        r = t | u;          // combine halfword minimums
+    #endif
+
+        return r;
+    }
+
+    // 4
+
+    static __device__ __forceinline__ unsigned int vadd4(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vadd4.u32.u32.u32.sat %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #elif __CUDA_ARCH__ >= 200
+        asm("vadd.u32.u32.u32.sat %0.b0, %1.b0, %2.b0, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+        asm("vadd.u32.u32.u32.sat %0.b1, %1.b1, %2.b1, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+        asm("vadd.u32.u32.u32.sat %0.b2, %1.b2, %2.b2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+        asm("vadd.u32.u32.u32.sat %0.b3, %1.b3, %2.b3, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        unsigned int s, t;
+        s = a ^ b;          // sum bits
+        r = a & 0x7f7f7f7f; // clear msbs
+        t = b & 0x7f7f7f7f; // clear msbs
+        s = s & 0x80808080; // msb sum bits
+        r = r + t;          // add without msbs, record carry-out in msbs
+        r = r ^ s;          // sum of msb sum and carry-in bits, w/o carry-out
+    #endif /* __CUDA_ARCH__ >= 300 */
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vsub4(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vsub4.u32.u32.u32.sat %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #elif __CUDA_ARCH__ >= 200
+        asm("vsub.u32.u32.u32.sat %0.b0, %1.b0, %2.b0, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+        asm("vsub.u32.u32.u32.sat %0.b1, %1.b1, %2.b1, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+        asm("vsub.u32.u32.u32.sat %0.b2, %1.b2, %2.b2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+        asm("vsub.u32.u32.u32.sat %0.b3, %1.b3, %2.b3, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        unsigned int s, t;
+        s = a ^ ~b;         // inverted sum bits
+        r = a | 0x80808080; // set msbs
+        t = b & 0x7f7f7f7f; // clear msbs
+        s = s & 0x80808080; // inverted msb sum bits
+        r = r - t;          // subtract w/o msbs, record inverted borrows in msb
+        r = r ^ s;          // combine inverted msb sum bits and borrows
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vavg4(unsigned int a, unsigned int b)
+    {
+        unsigned int r, s;
+
+        // HAKMEM #23: a + b = 2 * (a & b) + (a ^ b) ==>
+        // (a + b) / 2 = (a & b) + ((a ^ b) >> 1)
+        s = a ^ b;
+        r = a & b;
+        s = s & 0xfefefefe; // ensure following shift doesn't cross byte boundaries
+        s = s >> 1;
+        s = r + s;
+
+        return s;
+    }
+
+    static __device__ __forceinline__ unsigned int vavrg4(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vavrg4.u32.u32.u32 %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        // HAKMEM #23: a + b = 2 * (a | b) - (a ^ b) ==>
+        // (a + b + 1) / 2 = (a | b) - ((a ^ b) >> 1)
+        unsigned int c;
+        c = a ^ b;
+        r = a | b;
+        c = c & 0xfefefefe; // ensure following shift doesn't cross byte boundaries
+        c = c >> 1;
+        r = r - c;
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vseteq4(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vset4.u32.u32.eq %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        // inspired by Alan Mycroft's null-byte detection algorithm:
+        // null_byte(x) = ((x - 0x01010101) & (~x & 0x80808080))
+        unsigned int c;
+        r = a ^ b;          // 0x00 if a == b
+        c = r | 0x80808080; // set msbs, to catch carry out
+        r = r ^ c;          // extract msbs, msb = 1 if r < 0x80
+        c = c - 0x01010101; // msb = 0, if r was 0x00 or 0x80
+        c = r & ~c;         // msb = 1, if r was 0x00
+        r = c >> 7;         // convert to bool
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vcmpeq4(unsigned int a, unsigned int b)
+    {
+        unsigned int r, t;
+
+    #if __CUDA_ARCH__ >= 300
+        r = vseteq4(a, b);
+        t = r << 8;         // convert bool
+        r = t - r;          //  to mask
+    #else
+        // inspired by Alan Mycroft's null-byte detection algorithm:
+        // null_byte(x) = ((x - 0x01010101) & (~x & 0x80808080))
+        t = a ^ b;          // 0x00 if a == b
+        r = t | 0x80808080; // set msbs, to catch carry out
+        t = t ^ r;          // extract msbs, msb = 1 if t < 0x80
+        r = r - 0x01010101; // msb = 0, if t was 0x00 or 0x80
+        r = t & ~r;         // msb = 1, if t was 0x00
+        t = r >> 7;         // build mask
+        t = r - t;          //  from
+        r = t | r;          //   msbs
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vsetle4(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vset4.u32.u32.le %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        unsigned int c;
+        asm("not.b32 %0, %0;" : "+r"(a));
+        c = vavrg4(a, b);   // (b + ~a + 1) / 2 = (b - a) / 2
+        c = c & 0x80808080; // msb = carry-outs
+        r = c >> 7;         // convert to bool
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vcmple4(unsigned int a, unsigned int b)
+    {
+        unsigned int r, c;
+
+    #if __CUDA_ARCH__ >= 300
+        r = vsetle4(a, b);
+        c = r << 8;         // convert bool
+        r = c - r;          //  to mask
+    #else
+        asm("not.b32 %0, %0;" : "+r"(a));
+        c = vavrg4(a, b);   // (b + ~a + 1) / 2 = (b - a) / 2
+        c = c & 0x80808080; // msbs = carry-outs
+        r = c >> 7;         // convert
+        r = c - r;          //  msbs to
+        r = c | r;          //   mask
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vsetlt4(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vset4.u32.u32.lt %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        unsigned int c;
+        asm("not.b32 %0, %0;" : "+r"(a));
+        c = vavg4(a, b);    // (b + ~a) / 2 = (b - a) / 2 [rounded down]
+        c = c & 0x80808080; // msb = carry-outs
+        r = c >> 7;         // convert to bool
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vcmplt4(unsigned int a, unsigned int b)
+    {
+        unsigned int r, c;
+
+    #if __CUDA_ARCH__ >= 300
+        r = vsetlt4(a, b);
+        c = r << 8;         // convert bool
+        r = c - r;          //  to mask
+    #else
+        asm("not.b32 %0, %0;" : "+r"(a));
+        c = vavg4(a, b);    // (b + ~a) / 2 = (b - a) / 2 [rounded down]
+        c = c & 0x80808080; // msbs = carry-outs
+        r = c >> 7;         // convert
+        r = c - r;          //  msbs to
+        r = c | r;          //   mask
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vsetge4(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vset4.u32.u32.ge %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        unsigned int c;
+        asm("not.b32 %0, %0;" : "+r"(b));
+        c = vavrg4(a, b);   // (a + ~b + 1) / 2 = (a - b) / 2
+        c = c & 0x80808080; // msb = carry-outs
+        r = c >> 7;         // convert to bool
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vcmpge4(unsigned int a, unsigned int b)
+    {
+        unsigned int r, s;
+
+    #if __CUDA_ARCH__ >= 300
+        r = vsetge4(a, b);
+        s = r << 8;         // convert bool
+        r = s - r;          //  to mask
+    #else
+        asm ("not.b32 %0,%0;" : "+r"(b));
+        r = vavrg4 (a, b);  // (a + ~b + 1) / 2 = (a - b) / 2
+        r = r & 0x80808080; // msb = carry-outs
+        s = r >> 7;         // build mask
+        s = r - s;          //  from
+        r = s | r;          //   msbs
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vsetgt4(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vset4.u32.u32.gt %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        unsigned int c;
+        asm("not.b32 %0, %0;" : "+r"(b));
+        c = vavg4(a, b);    // (a + ~b) / 2 = (a - b) / 2 [rounded down]
+        c = c & 0x80808080; // msb = carry-outs
+        r = c >> 7;         // convert to bool
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vcmpgt4(unsigned int a, unsigned int b)
+    {
+        unsigned int r, c;
+
+    #if __CUDA_ARCH__ >= 300
+        r = vsetgt4(a, b);
+        c = r << 8;         // convert bool
+        r = c - r;          //  to mask
+    #else
+        asm("not.b32 %0, %0;" : "+r"(b));
+        c = vavg4(a, b);    // (a + ~b) / 2 = (a - b) / 2 [rounded down]
+        c = c & 0x80808080; // msb = carry-outs
+        r = c >> 7;         // convert
+        r = c - r;          //  msbs to
+        r = c | r;          //   mask
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vsetne4(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vset4.u32.u32.ne %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        // inspired by Alan Mycroft's null-byte detection algorithm:
+        // null_byte(x) = ((x - 0x01010101) & (~x & 0x80808080))
+        unsigned int c;
+        r = a ^ b;          // 0x00 if a == b
+        c = r | 0x80808080; // set msbs, to catch carry out
+        c = c - 0x01010101; // msb = 0, if r was 0x00 or 0x80
+        c = r | c;          // msb = 1, if r was not 0x00
+        c = c & 0x80808080; // extract msbs
+        r = c >> 7;         // convert to bool
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vcmpne4(unsigned int a, unsigned int b)
+    {
+        unsigned int r, c;
+
+    #if __CUDA_ARCH__ >= 300
+        r = vsetne4(a, b);
+        c = r << 8;         // convert bool
+        r = c - r;          //  to mask
+    #else
+        // inspired by Alan Mycroft's null-byte detection algorithm:
+        // null_byte(x) = ((x - 0x01010101) & (~x & 0x80808080))
+        r = a ^ b;          // 0x00 if a == b
+        c = r | 0x80808080; // set msbs, to catch carry out
+        c = c - 0x01010101; // msb = 0, if r was 0x00 or 0x80
+        c = r | c;          // msb = 1, if r was not 0x00
+        c = c & 0x80808080; // extract msbs
+        r = c >> 7;         // convert
+        r = c - r;          //  msbs to
+        r = c | r;          //   mask
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vabsdiff4(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vabsdiff4.u32.u32.u32.sat %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #elif __CUDA_ARCH__ >= 200
+        asm("vabsdiff.u32.u32.u32.sat %0.b0, %1.b0, %2.b0, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+        asm("vabsdiff.u32.u32.u32.sat %0.b1, %1.b1, %2.b1, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+        asm("vabsdiff.u32.u32.u32.sat %0.b2, %1.b2, %2.b2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+        asm("vabsdiff.u32.u32.u32.sat %0.b3, %1.b3, %2.b3, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        unsigned int s;
+        s = vcmpge4(a, b);  // mask = 0xff if a >= b
+        r = a ^ b;          //
+        s = (r &  s) ^ b;   // select a when a >= b, else select b => max(a,b)
+        r = s ^ r;          // select a when b >= a, else select b => min(a,b)
+        r = s - r;          // |a - b| = max(a,b) - min(a,b);
+    #endif
+
+        return r;
+    }
+
+    static __device__ __forceinline__ unsigned int vmax4(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vmax4.u32.u32.u32 %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #elif __CUDA_ARCH__ >= 200
+        asm("vmax.u32.u32.u32 %0.b0, %1.b0, %2.b0, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+        asm("vmax.u32.u32.u32 %0.b1, %1.b1, %2.b1, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+        asm("vmax.u32.u32.u32 %0.b2, %1.b2, %2.b2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+        asm("vmax.u32.u32.u32 %0.b3, %1.b3, %2.b3, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        unsigned int s;
+        s = vcmpge4(a, b);  // mask = 0xff if a >= b
+        r = a & s;          // select a when b >= a
+        s = b & ~s;         // select b when b < a
+        r = r | s;          // combine byte selections
+    #endif
+
+        return r;           // byte-wise unsigned maximum
+    }
+
+    static __device__ __forceinline__ unsigned int vmin4(unsigned int a, unsigned int b)
+    {
+        unsigned int r = 0;
+
+    #if __CUDA_ARCH__ >= 300
+        asm("vmin4.u32.u32.u32 %0, %1, %2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #elif __CUDA_ARCH__ >= 200
+        asm("vmin.u32.u32.u32 %0.b0, %1.b0, %2.b0, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+        asm("vmin.u32.u32.u32 %0.b1, %1.b1, %2.b1, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+        asm("vmin.u32.u32.u32 %0.b2, %1.b2, %2.b2, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+        asm("vmin.u32.u32.u32 %0.b3, %1.b3, %2.b3, %3;" : "=r"(r) : "r"(a), "r"(b), "r"(r));
+    #else
+        unsigned int s;
+        s = vcmpge4(b, a);  // mask = 0xff if a >= b
+        r = a & s;          // select a when b >= a
+        s = b & ~s;         // select b when b < a
+        r = r | s;          // combine byte selections
+    #endif
+
+        return r;
+    }
+}}}
+
+//! @endcond
+
+#endif // OPENCV_CUDA_SIMD_FUNCTIONS_HPP

model_zoo/official/lite/app/src/main/cpp/include/opencv2/core/cuda/transform.hpp

@@ -0,0 +1,75 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CUDA_TRANSFORM_HPP
+#define OPENCV_CUDA_TRANSFORM_HPP
+
+#include "common.hpp"
+#include "utility.hpp"
+#include "detail/transform_detail.hpp"
+
+/** @file
+ * @deprecated Use @ref cudev instead.
+ */
+
+//! @cond IGNORED
+
+namespace cv { namespace cuda { namespace device
+{
+    template <typename T, typename D, typename UnOp, typename Mask>
+    static inline void transform(PtrStepSz<T> src, PtrStepSz<D> dst, UnOp op, const Mask& mask, cudaStream_t stream)
+    {
+        typedef TransformFunctorTraits<UnOp> ft;
+        transform_detail::TransformDispatcher<VecTraits<T>::cn == 1 && VecTraits<D>::cn == 1 && ft::smart_shift != 1>::call(src, dst, op, mask, stream);
+    }
+
+    template <typename T1, typename T2, typename D, typename BinOp, typename Mask>
+    static inline void transform(PtrStepSz<T1> src1, PtrStepSz<T2> src2, PtrStepSz<D> dst, BinOp op, const Mask& mask, cudaStream_t stream)
+    {
+        typedef TransformFunctorTraits<BinOp> ft;
+        transform_detail::TransformDispatcher<VecTraits<T1>::cn == 1 && VecTraits<T2>::cn == 1 && VecTraits<D>::cn == 1 && ft::smart_shift != 1>::call(src1, src2, dst, op, mask, stream);
+    }
+}}}
+
+//! @endcond
+
+#endif // OPENCV_CUDA_TRANSFORM_HPP

model_zoo/official/lite/app/src/main/cpp/include/opencv2/core/cuda/type_traits.hpp

@@ -0,0 +1,90 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CUDA_TYPE_TRAITS_HPP
+#define OPENCV_CUDA_TYPE_TRAITS_HPP
+
+#include "detail/type_traits_detail.hpp"
+
+/** @file
+ * @deprecated Use @ref cudev instead.
+ */
+
+//! @cond IGNORED
+
+namespace cv { namespace cuda { namespace device
+{
+    template <typename T> struct IsSimpleParameter
+    {
+        enum {value = type_traits_detail::IsIntegral<T>::value || type_traits_detail::IsFloat<T>::value ||
+            type_traits_detail::PointerTraits<typename type_traits_detail::ReferenceTraits<T>::type>::value};
+    };
+
+    template <typename T> struct TypeTraits
+    {
+        typedef typename type_traits_detail::UnConst<T>::type                                                NonConstType;
+        typedef typename type_traits_detail::UnVolatile<T>::type                                             NonVolatileType;
+        typedef typename type_traits_detail::UnVolatile<typename type_traits_detail::UnConst<T>::type>::type UnqualifiedType;
+        typedef typename type_traits_detail::PointerTraits<UnqualifiedType>::type                            PointeeType;
+        typedef typename type_traits_detail::ReferenceTraits<T>::type                                        ReferredType;
+
+        enum { isConst          = type_traits_detail::UnConst<T>::value };
+        enum { isVolatile       = type_traits_detail::UnVolatile<T>::value };
+
+        enum { isReference      = type_traits_detail::ReferenceTraits<UnqualifiedType>::value };
+        enum { isPointer        = type_traits_detail::PointerTraits<typename type_traits_detail::ReferenceTraits<UnqualifiedType>::type>::value };
+
+        enum { isUnsignedInt    = type_traits_detail::IsUnsignedIntegral<UnqualifiedType>::value };
+        enum { isSignedInt      = type_traits_detail::IsSignedIntergral<UnqualifiedType>::value };
+        enum { isIntegral       = type_traits_detail::IsIntegral<UnqualifiedType>::value };
+        enum { isFloat          = type_traits_detail::IsFloat<UnqualifiedType>::value };
+        enum { isArith          = isIntegral || isFloat };
+        enum { isVec            = type_traits_detail::IsVec<UnqualifiedType>::value };
+
+        typedef typename type_traits_detail::Select<IsSimpleParameter<UnqualifiedType>::value,
+            T, typename type_traits_detail::AddParameterType<T>::type>::type ParameterType;
+    };
+}}}
+
+//! @endcond
+
+#endif // OPENCV_CUDA_TYPE_TRAITS_HPP

model_zoo/official/lite/app/src/main/cpp/include/opencv2/core/cuda/utility.hpp

@@ -0,0 +1,230 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CUDA_UTILITY_HPP
+#define OPENCV_CUDA_UTILITY_HPP
+
+#include "saturate_cast.hpp"
+#include "datamov_utils.hpp"
+
+/** @file
+ * @deprecated Use @ref cudev instead.
+ */
+
+//! @cond IGNORED
+
+namespace cv { namespace cuda { namespace device
+{
+    struct CV_EXPORTS ThrustAllocator
+    {
+        typedef uchar value_type;
+        virtual ~ThrustAllocator();
+        virtual __device__ __host__ uchar* allocate(size_t numBytes) = 0;
+        virtual __device__ __host__ void deallocate(uchar* ptr, size_t numBytes) = 0;
+        static ThrustAllocator& getAllocator();
+        static void setAllocator(ThrustAllocator* allocator);
+    };
+    #define OPENCV_CUDA_LOG_WARP_SIZE        (5)
+    #define OPENCV_CUDA_WARP_SIZE            (1 << OPENCV_CUDA_LOG_WARP_SIZE)
+    #define OPENCV_CUDA_LOG_MEM_BANKS        ((__CUDA_ARCH__ >= 200) ? 5 : 4) // 32 banks on fermi, 16 on tesla
+    #define OPENCV_CUDA_MEM_BANKS            (1 << OPENCV_CUDA_LOG_MEM_BANKS)
+
+    ///////////////////////////////////////////////////////////////////////////////
+    // swap
+
+    template <typename T> void __device__ __host__ __forceinline__ swap(T& a, T& b)
+    {
+        const T temp = a;
+        a = b;
+        b = temp;
+    }
+
+    ///////////////////////////////////////////////////////////////////////////////
+    // Mask Reader
+
+    struct SingleMask
+    {
+        explicit __host__ __device__ __forceinline__ SingleMask(PtrStepb mask_) : mask(mask_) {}
+        __host__ __device__ __forceinline__ SingleMask(const SingleMask& mask_): mask(mask_.mask){}
+
+        __device__ __forceinline__ bool operator()(int y, int x) const
+        {
+            return mask.ptr(y)[x] != 0;
+        }
+
+        PtrStepb mask;
+    };
+
+    struct SingleMaskChannels
+    {
+        __host__ __device__ __forceinline__ SingleMaskChannels(PtrStepb mask_, int channels_)
+        : mask(mask_), channels(channels_) {}
+        __host__ __device__ __forceinline__ SingleMaskChannels(const SingleMaskChannels& mask_)
+            :mask(mask_.mask), channels(mask_.channels){}
+
+        __device__ __forceinline__ bool operator()(int y, int x) const
+        {
+            return mask.ptr(y)[x / channels] != 0;
+        }
+
+        PtrStepb mask;
+        int channels;
+    };
+
+    struct MaskCollection
+    {
+        explicit __host__ __device__ __forceinline__ MaskCollection(PtrStepb* maskCollection_)
+            : maskCollection(maskCollection_) {}
+
+        __device__ __forceinline__ MaskCollection(const MaskCollection& masks_)
+            : maskCollection(masks_.maskCollection), curMask(masks_.curMask){}
+
+        __device__ __forceinline__ void next()
+        {
+            curMask = *maskCollection++;
+        }
+        __device__ __forceinline__ void setMask(int z)
+        {
+            curMask = maskCollection[z];
+        }
+
+        __device__ __forceinline__ bool operator()(int y, int x) const
+        {
+            uchar val;
+            return curMask.data == 0 || (ForceGlob<uchar>::Load(curMask.ptr(y), x, val), (val != 0));
+        }
+
+        const PtrStepb* maskCollection;
+        PtrStepb curMask;
+    };
+
+    struct WithOutMask
+    {
+        __host__ __device__ __forceinline__ WithOutMask(){}
+        __host__ __device__ __forceinline__ WithOutMask(const WithOutMask&){}
+
+        __device__ __forceinline__ void next() const
+        {
+        }
+        __device__ __forceinline__ void setMask(int) const
+        {
+        }
+
+        __device__ __forceinline__ bool operator()(int, int) const
+        {
+            return true;
+        }
+
+        __device__ __forceinline__ bool operator()(int, int, int) const
+        {
+            return true;
+        }
+
+        static __device__ __forceinline__ bool check(int, int)
+        {
+            return true;
+        }
+
+        static __device__ __forceinline__ bool check(int, int, int)
+        {
+            return true;
+        }
+    };
+
+    ///////////////////////////////////////////////////////////////////////////////
+    // Solve linear system
+
+    // solve 2x2 linear system Ax=b
+    template <typename T> __device__ __forceinline__ bool solve2x2(const T A[2][2], const T b[2], T x[2])
+    {
+        T det = A[0][0] * A[1][1] - A[1][0] * A[0][1];
+
+        if (det != 0)
+        {
+            double invdet = 1.0 / det;
+
+            x[0] = saturate_cast<T>(invdet * (b[0] * A[1][1] - b[1] * A[0][1]));
+
+            x[1] = saturate_cast<T>(invdet * (A[0][0] * b[1] - A[1][0] * b[0]));
+
+            return true;
+        }
+
+        return false;
+    }
+
+    // solve 3x3 linear system Ax=b
+    template <typename T> __device__ __forceinline__ bool solve3x3(const T A[3][3], const T b[3], T x[3])
+    {
+        T det = A[0][0] * (A[1][1] * A[2][2] - A[1][2] * A[2][1])
+              - A[0][1] * (A[1][0] * A[2][2] - A[1][2] * A[2][0])
+              + A[0][2] * (A[1][0] * A[2][1] - A[1][1] * A[2][0]);
+
+        if (det != 0)
+        {
+            double invdet = 1.0 / det;
+
+            x[0] = saturate_cast<T>(invdet *
+                (b[0]    * (A[1][1] * A[2][2] - A[1][2] * A[2][1]) -
+                 A[0][1] * (b[1]    * A[2][2] - A[1][2] * b[2]   ) +
+                 A[0][2] * (b[1]    * A[2][1] - A[1][1] * b[2]   )));
+
+            x[1] = saturate_cast<T>(invdet *
+                (A[0][0] * (b[1]    * A[2][2] - A[1][2] * b[2]   ) -
+                 b[0]    * (A[1][0] * A[2][2] - A[1][2] * A[2][0]) +
+                 A[0][2] * (A[1][0] * b[2]    - b[1]    * A[2][0])));
+
+            x[2] = saturate_cast<T>(invdet *
+                (A[0][0] * (A[1][1] * b[2]    - b[1]    * A[2][1]) -
+                 A[0][1] * (A[1][0] * b[2]    - b[1]    * A[2][0]) +
+                 b[0]    * (A[1][0] * A[2][1] - A[1][1] * A[2][0])));
+
+            return true;
+        }
+
+        return false;
+    }
+}}} // namespace cv { namespace cuda { namespace cudev
+
+//! @endcond
+
+#endif // OPENCV_CUDA_UTILITY_HPP

model_zoo/official/lite/app/src/main/cpp/include/opencv2/core/cuda/vec_distance.hpp

@@ -0,0 +1,232 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CUDA_VEC_DISTANCE_HPP
+#define OPENCV_CUDA_VEC_DISTANCE_HPP
+
+#include "reduce.hpp"
+#include "functional.hpp"
+#include "detail/vec_distance_detail.hpp"
+
+/** @file
+ * @deprecated Use @ref cudev instead.
+ */
+
+//! @cond IGNORED
+
+namespace cv { namespace cuda { namespace device
+{
+    template <typename T> struct L1Dist
+    {
+        typedef int value_type;
+        typedef int result_type;
+
+        __device__ __forceinline__ L1Dist() : mySum(0) {}
+
+        __device__ __forceinline__ void reduceIter(int val1, int val2)
+        {
+            mySum = __sad(val1, val2, mySum);
+        }
+
+        template <int THREAD_DIM> __device__ __forceinline__ void reduceAll(int* smem, int tid)
+        {
+            reduce<THREAD_DIM>(smem, mySum, tid, plus<int>());
+        }
+
+        __device__ __forceinline__ operator int() const
+        {
+            return mySum;
+        }
+
+        int mySum;
+    };
+    template <> struct L1Dist<float>
+    {
+        typedef float value_type;
+        typedef float result_type;
+
+        __device__ __forceinline__ L1Dist() : mySum(0.0f) {}
+
+        __device__ __forceinline__ void reduceIter(float val1, float val2)
+        {
+            mySum += ::fabs(val1 - val2);
+        }
+
+        template <int THREAD_DIM> __device__ __forceinline__ void reduceAll(float* smem, int tid)
+        {
+            reduce<THREAD_DIM>(smem, mySum, tid, plus<float>());
+        }
+
+        __device__ __forceinline__ operator float() const
+        {
+            return mySum;
+        }
+
+        float mySum;
+    };
+
+    struct L2Dist
+    {
+        typedef float value_type;
+        typedef float result_type;
+
+        __device__ __forceinline__ L2Dist() : mySum(0.0f) {}
+
+        __device__ __forceinline__ void reduceIter(float val1, float val2)
+        {
+            float reg = val1 - val2;
+            mySum += reg * reg;
+        }
+
+        template <int THREAD_DIM> __device__ __forceinline__ void reduceAll(float* smem, int tid)
+        {
+            reduce<THREAD_DIM>(smem, mySum, tid, plus<float>());
+        }
+
+        __device__ __forceinline__ operator float() const
+        {
+            return sqrtf(mySum);
+        }
+
+        float mySum;
+    };
+
+    struct HammingDist
+    {
+        typedef int value_type;
+        typedef int result_type;
+
+        __device__ __forceinline__ HammingDist() : mySum(0) {}
+
+        __device__ __forceinline__ void reduceIter(int val1, int val2)
+        {
+            mySum += __popc(val1 ^ val2);
+        }
+
+        template <int THREAD_DIM> __device__ __forceinline__ void reduceAll(int* smem, int tid)
+        {
+            reduce<THREAD_DIM>(smem, mySum, tid, plus<int>());
+        }
+
+        __device__ __forceinline__ operator int() const
+        {
+            return mySum;
+        }
+
+        int mySum;
+    };
+
+    // calc distance between two vectors in global memory
+    template <int THREAD_DIM, typename Dist, typename T1, typename T2>
+    __device__ void calcVecDiffGlobal(const T1* vec1, const T2* vec2, int len, Dist& dist, typename Dist::result_type* smem, int tid)
+    {
+        for (int i = tid; i < len; i += THREAD_DIM)
+        {
+            T1 val1;
+            ForceGlob<T1>::Load(vec1, i, val1);
+
+            T2 val2;
+            ForceGlob<T2>::Load(vec2, i, val2);
+
+            dist.reduceIter(val1, val2);
+        }
+
+        dist.reduceAll<THREAD_DIM>(smem, tid);
+    }
+
+    // calc distance between two vectors, first vector is cached in register or shared memory, second vector is in global memory
+    template <int THREAD_DIM, int MAX_LEN, bool LEN_EQ_MAX_LEN, typename Dist, typename T1, typename T2>
+    __device__ __forceinline__ void calcVecDiffCached(const T1* vecCached, const T2* vecGlob, int len, Dist& dist, typename Dist::result_type* smem, int tid)
+    {
+        vec_distance_detail::VecDiffCachedCalculator<THREAD_DIM, MAX_LEN, LEN_EQ_MAX_LEN>::calc(vecCached, vecGlob, len, dist, tid);
+
+        dist.reduceAll<THREAD_DIM>(smem, tid);
+    }
+
+    // calc distance between two vectors in global memory
+    template <int THREAD_DIM, typename T1> struct VecDiffGlobal
+    {
+        explicit __device__ __forceinline__ VecDiffGlobal(const T1* vec1_, int = 0, void* = 0, int = 0, int = 0)
+        {
+            vec1 = vec1_;
+        }
+
+        template <typename T2, typename Dist>
+        __device__ __forceinline__ void calc(const T2* vec2, int len, Dist& dist, typename Dist::result_type* smem, int tid) const
+        {
+            calcVecDiffGlobal<THREAD_DIM>(vec1, vec2, len, dist, smem, tid);
+        }
+
+        const T1* vec1;
+    };
+
+    // calc distance between two vectors, first vector is cached in register memory, second vector is in global memory
+    template <int THREAD_DIM, int MAX_LEN, bool LEN_EQ_MAX_LEN, typename U> struct VecDiffCachedRegister
+    {
+        template <typename T1> __device__ __forceinline__ VecDiffCachedRegister(const T1* vec1, int len, U* smem, int glob_tid, int tid)
+        {
+            if (glob_tid < len)
+                smem[glob_tid] = vec1[glob_tid];
+            __syncthreads();
+
+            U* vec1ValsPtr = vec1Vals;
+
+            #pragma unroll
+            for (int i = tid; i < MAX_LEN; i += THREAD_DIM)
+                *vec1ValsPtr++ = smem[i];
+
+            __syncthreads();
+        }
+
+        template <typename T2, typename Dist>
+        __device__ __forceinline__ void calc(const T2* vec2, int len, Dist& dist, typename Dist::result_type* smem, int tid) const
+        {
+            calcVecDiffCached<THREAD_DIM, MAX_LEN, LEN_EQ_MAX_LEN>(vec1Vals, vec2, len, dist, smem, tid);
+        }
+
+        U vec1Vals[MAX_LEN / THREAD_DIM];
+    };
+}}} // namespace cv { namespace cuda { namespace cudev
+
+//! @endcond
+
+#endif // OPENCV_CUDA_VEC_DISTANCE_HPP

model_zoo/official/lite/app/src/main/cpp/include/opencv2/core/cuda/vec_math.hpp

@@ -0,0 +1,923 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CUDA_VECMATH_HPP
+#define OPENCV_CUDA_VECMATH_HPP
+
+#include "vec_traits.hpp"
+#include "saturate_cast.hpp"
+
+/** @file
+ * @deprecated Use @ref cudev instead.
+ */
+
+//! @cond IGNORED
+
+namespace cv { namespace cuda { namespace device
+{
+
+// saturate_cast
+
+namespace vec_math_detail
+{
+    template <int cn, typename VecD> struct SatCastHelper;
+    template <typename VecD> struct SatCastHelper<1, VecD>
+    {
+        template <typename VecS> static __device__ __forceinline__ VecD cast(const VecS& v)
+        {
+            typedef typename VecTraits<VecD>::elem_type D;
+            return VecTraits<VecD>::make(saturate_cast<D>(v.x));
+        }
+    };
+    template <typename VecD> struct SatCastHelper<2, VecD>
+    {
+        template <typename VecS> static __device__ __forceinline__ VecD cast(const VecS& v)
+        {
+            typedef typename VecTraits<VecD>::elem_type D;
+            return VecTraits<VecD>::make(saturate_cast<D>(v.x), saturate_cast<D>(v.y));
+        }
+    };
+    template <typename VecD> struct SatCastHelper<3, VecD>
+    {
+        template <typename VecS> static __device__ __forceinline__ VecD cast(const VecS& v)
+        {
+            typedef typename VecTraits<VecD>::elem_type D;
+            return VecTraits<VecD>::make(saturate_cast<D>(v.x), saturate_cast<D>(v.y), saturate_cast<D>(v.z));
+        }
+    };
+    template <typename VecD> struct SatCastHelper<4, VecD>
+    {
+        template <typename VecS> static __device__ __forceinline__ VecD cast(const VecS& v)
+        {
+            typedef typename VecTraits<VecD>::elem_type D;
+            return VecTraits<VecD>::make(saturate_cast<D>(v.x), saturate_cast<D>(v.y), saturate_cast<D>(v.z), saturate_cast<D>(v.w));
+        }
+    };
+
+    template <typename VecD, typename VecS> static __device__ __forceinline__ VecD saturate_cast_helper(const VecS& v)
+    {
+        return SatCastHelper<VecTraits<VecD>::cn, VecD>::cast(v);
+    }
+}
+
+template<typename T> static __device__ __forceinline__ T saturate_cast(const uchar1& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const char1& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const ushort1& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const short1& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const uint1& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const int1& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const float1& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const double1& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+
+template<typename T> static __device__ __forceinline__ T saturate_cast(const uchar2& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const char2& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const ushort2& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const short2& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const uint2& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const int2& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const float2& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const double2& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+
+template<typename T> static __device__ __forceinline__ T saturate_cast(const uchar3& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const char3& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const ushort3& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const short3& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const uint3& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const int3& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const float3& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const double3& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+
+template<typename T> static __device__ __forceinline__ T saturate_cast(const uchar4& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const char4& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const ushort4& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const short4& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const uint4& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const int4& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const float4& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+template<typename T> static __device__ __forceinline__ T saturate_cast(const double4& v) {return vec_math_detail::saturate_cast_helper<T>(v);}
+
+// unary operators
+
+#define CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(op, input_type, output_type) \
+    __device__ __forceinline__ output_type ## 1 operator op(const input_type ## 1 & a) \
+    { \
+        return VecTraits<output_type ## 1>::make(op (a.x)); \
+    } \
+    __device__ __forceinline__ output_type ## 2 operator op(const input_type ## 2 & a) \
+    { \
+        return VecTraits<output_type ## 2>::make(op (a.x), op (a.y)); \
+    } \
+    __device__ __forceinline__ output_type ## 3 operator op(const input_type ## 3 & a) \
+    { \
+        return VecTraits<output_type ## 3>::make(op (a.x), op (a.y), op (a.z)); \
+    } \
+    __device__ __forceinline__ output_type ## 4 operator op(const input_type ## 4 & a) \
+    { \
+        return VecTraits<output_type ## 4>::make(op (a.x), op (a.y), op (a.z), op (a.w)); \
+    }
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(-, char, char)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(-, short, short)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(-, int, int)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(-, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(-, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(!, uchar, uchar)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(!, char, uchar)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(!, ushort, uchar)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(!, short, uchar)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(!, int, uchar)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(!, uint, uchar)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(!, float, uchar)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(!, double, uchar)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(~, uchar, uchar)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(~, char, char)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(~, ushort, ushort)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(~, short, short)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(~, int, int)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_OP(~, uint, uint)
+
+#undef CV_CUDEV_IMPLEMENT_VEC_UNARY_OP
+
+// unary functions
+
+#define CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(func_name, func, input_type, output_type) \
+    __device__ __forceinline__ output_type ## 1 func_name(const input_type ## 1 & a) \
+    { \
+        return VecTraits<output_type ## 1>::make(func (a.x)); \
+    } \
+    __device__ __forceinline__ output_type ## 2 func_name(const input_type ## 2 & a) \
+    { \
+        return VecTraits<output_type ## 2>::make(func (a.x), func (a.y)); \
+    } \
+    __device__ __forceinline__ output_type ## 3 func_name(const input_type ## 3 & a) \
+    { \
+        return VecTraits<output_type ## 3>::make(func (a.x), func (a.y), func (a.z)); \
+    } \
+    __device__ __forceinline__ output_type ## 4 func_name(const input_type ## 4 & a) \
+    { \
+        return VecTraits<output_type ## 4>::make(func (a.x), func (a.y), func (a.z), func (a.w)); \
+    }
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(abs, ::fabsf, float, float)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sqrt, ::sqrtf, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sqrt, ::sqrtf, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sqrt, ::sqrtf, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sqrt, ::sqrtf, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sqrt, ::sqrtf, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sqrt, ::sqrtf, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sqrt, ::sqrtf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sqrt, ::sqrt, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp, ::expf, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp, ::expf, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp, ::expf, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp, ::expf, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp, ::expf, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp, ::expf, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp, ::expf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp, ::exp, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp2, ::exp2f, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp2, ::exp2f, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp2, ::exp2f, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp2, ::exp2f, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp2, ::exp2f, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp2, ::exp2f, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp2, ::exp2f, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp2, ::exp2, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp10, ::exp10f, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp10, ::exp10f, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp10, ::exp10f, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp10, ::exp10f, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp10, ::exp10f, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp10, ::exp10f, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp10, ::exp10f, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(exp10, ::exp10, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log, ::logf, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log, ::logf, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log, ::logf, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log, ::logf, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log, ::logf, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log, ::logf, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log, ::logf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log, ::log, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log2, ::log2f, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log2, ::log2f, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log2, ::log2f, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log2, ::log2f, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log2, ::log2f, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log2, ::log2f, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log2, ::log2f, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log2, ::log2, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log10, ::log10f, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log10, ::log10f, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log10, ::log10f, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log10, ::log10f, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log10, ::log10f, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log10, ::log10f, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log10, ::log10f, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(log10, ::log10, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sin, ::sinf, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sin, ::sinf, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sin, ::sinf, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sin, ::sinf, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sin, ::sinf, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sin, ::sinf, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sin, ::sinf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sin, ::sin, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(cos, ::cosf, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(cos, ::cosf, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(cos, ::cosf, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(cos, ::cosf, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(cos, ::cosf, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(cos, ::cosf, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(cos, ::cosf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(cos, ::cos, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(tan, ::tanf, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(tan, ::tanf, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(tan, ::tanf, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(tan, ::tanf, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(tan, ::tanf, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(tan, ::tanf, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(tan, ::tanf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(tan, ::tan, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(asin, ::asinf, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(asin, ::asinf, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(asin, ::asinf, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(asin, ::asinf, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(asin, ::asinf, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(asin, ::asinf, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(asin, ::asinf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(asin, ::asin, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(acos, ::acosf, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(acos, ::acosf, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(acos, ::acosf, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(acos, ::acosf, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(acos, ::acosf, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(acos, ::acosf, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(acos, ::acosf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(acos, ::acos, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(atan, ::atanf, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(atan, ::atanf, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(atan, ::atanf, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(atan, ::atanf, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(atan, ::atanf, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(atan, ::atanf, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(atan, ::atanf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(atan, ::atan, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sinh, ::sinhf, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sinh, ::sinhf, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sinh, ::sinhf, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sinh, ::sinhf, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sinh, ::sinhf, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sinh, ::sinhf, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sinh, ::sinhf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(sinh, ::sinh, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(cosh, ::coshf, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(cosh, ::coshf, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(cosh, ::coshf, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(cosh, ::coshf, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(cosh, ::coshf, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(cosh, ::coshf, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(cosh, ::coshf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(cosh, ::cosh, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(tanh, ::tanhf, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(tanh, ::tanhf, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(tanh, ::tanhf, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(tanh, ::tanhf, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(tanh, ::tanhf, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(tanh, ::tanhf, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(tanh, ::tanhf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(tanh, ::tanh, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(asinh, ::asinhf, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(asinh, ::asinhf, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(asinh, ::asinhf, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(asinh, ::asinhf, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(asinh, ::asinhf, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(asinh, ::asinhf, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(asinh, ::asinhf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(asinh, ::asinh, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(acosh, ::acoshf, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(acosh, ::acoshf, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(acosh, ::acoshf, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(acosh, ::acoshf, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(acosh, ::acoshf, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(acosh, ::acoshf, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(acosh, ::acoshf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(acosh, ::acosh, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(atanh, ::atanhf, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(atanh, ::atanhf, char, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(atanh, ::atanhf, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(atanh, ::atanhf, short, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(atanh, ::atanhf, int, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(atanh, ::atanhf, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(atanh, ::atanhf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC(atanh, ::atanh, double, double)
+
+#undef CV_CUDEV_IMPLEMENT_VEC_UNARY_FUNC
+
+// binary operators (vec & vec)
+
+#define CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(op, input_type, output_type) \
+    __device__ __forceinline__ output_type ## 1 operator op(const input_type ## 1 & a, const input_type ## 1 & b) \
+    { \
+        return VecTraits<output_type ## 1>::make(a.x op b.x); \
+    } \
+    __device__ __forceinline__ output_type ## 2 operator op(const input_type ## 2 & a, const input_type ## 2 & b) \
+    { \
+        return VecTraits<output_type ## 2>::make(a.x op b.x, a.y op b.y); \
+    } \
+    __device__ __forceinline__ output_type ## 3 operator op(const input_type ## 3 & a, const input_type ## 3 & b) \
+    { \
+        return VecTraits<output_type ## 3>::make(a.x op b.x, a.y op b.y, a.z op b.z); \
+    } \
+    __device__ __forceinline__ output_type ## 4 operator op(const input_type ## 4 & a, const input_type ## 4 & b) \
+    { \
+        return VecTraits<output_type ## 4>::make(a.x op b.x, a.y op b.y, a.z op b.z, a.w op b.w); \
+    }
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(+, uchar, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(+, char, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(+, ushort, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(+, short, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(+, int, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(+, uint, uint)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(+, float, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(+, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(-, uchar, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(-, char, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(-, ushort, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(-, short, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(-, int, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(-, uint, uint)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(-, float, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(-, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(*, uchar, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(*, char, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(*, ushort, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(*, short, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(*, int, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(*, uint, uint)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(*, float, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(*, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(/, uchar, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(/, char, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(/, ushort, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(/, short, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(/, int, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(/, uint, uint)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(/, float, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(/, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(==, uchar, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(==, char, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(==, ushort, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(==, short, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(==, int, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(==, uint, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(==, float, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(==, double, uchar)
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(!=, uchar, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(!=, char, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(!=, ushort, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(!=, short, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(!=, int, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(!=, uint, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(!=, float, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(!=, double, uchar)
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(>, uchar, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(>, char, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(>, ushort, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(>, short, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(>, int, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(>, uint, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(>, float, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(>, double, uchar)
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(<, uchar, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(<, char, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(<, ushort, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(<, short, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(<, int, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(<, uint, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(<, float, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(<, double, uchar)
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(>=, uchar, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(>=, char, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(>=, ushort, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(>=, short, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(>=, int, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(>=, uint, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(>=, float, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(>=, double, uchar)
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(<=, uchar, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(<=, char, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(<=, ushort, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(<=, short, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(<=, int, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(<=, uint, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(<=, float, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(<=, double, uchar)
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(&&, uchar, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(&&, char, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(&&, ushort, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(&&, short, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(&&, int, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(&&, uint, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(&&, float, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(&&, double, uchar)
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(||, uchar, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(||, char, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(||, ushort, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(||, short, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(||, int, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(||, uint, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(||, float, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(||, double, uchar)
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(&, uchar, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(&, char, char)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(&, ushort, ushort)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(&, short, short)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(&, int, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(&, uint, uint)
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(|, uchar, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(|, char, char)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(|, ushort, ushort)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(|, short, short)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(|, int, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(|, uint, uint)
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(^, uchar, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(^, char, char)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(^, ushort, ushort)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(^, short, short)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(^, int, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_OP(^, uint, uint)
+
+#undef CV_CUDEV_IMPLEMENT_VEC_BINARY_OP
+
+// binary operators (vec & scalar)
+
+#define CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(op, input_type, scalar_type, output_type) \
+    __device__ __forceinline__ output_type ## 1 operator op(const input_type ## 1 & a, scalar_type s) \
+    { \
+        return VecTraits<output_type ## 1>::make(a.x op s); \
+    } \
+    __device__ __forceinline__ output_type ## 1 operator op(scalar_type s, const input_type ## 1 & b) \
+    { \
+        return VecTraits<output_type ## 1>::make(s op b.x); \
+    } \
+    __device__ __forceinline__ output_type ## 2 operator op(const input_type ## 2 & a, scalar_type s) \
+    { \
+        return VecTraits<output_type ## 2>::make(a.x op s, a.y op s); \
+    } \
+    __device__ __forceinline__ output_type ## 2 operator op(scalar_type s, const input_type ## 2 & b) \
+    { \
+        return VecTraits<output_type ## 2>::make(s op b.x, s op b.y); \
+    } \
+    __device__ __forceinline__ output_type ## 3 operator op(const input_type ## 3 & a, scalar_type s) \
+    { \
+        return VecTraits<output_type ## 3>::make(a.x op s, a.y op s, a.z op s); \
+    } \
+    __device__ __forceinline__ output_type ## 3 operator op(scalar_type s, const input_type ## 3 & b) \
+    { \
+        return VecTraits<output_type ## 3>::make(s op b.x, s op b.y, s op b.z); \
+    } \
+    __device__ __forceinline__ output_type ## 4 operator op(const input_type ## 4 & a, scalar_type s) \
+    { \
+        return VecTraits<output_type ## 4>::make(a.x op s, a.y op s, a.z op s, a.w op s); \
+    } \
+    __device__ __forceinline__ output_type ## 4 operator op(scalar_type s, const input_type ## 4 & b) \
+    { \
+        return VecTraits<output_type ## 4>::make(s op b.x, s op b.y, s op b.z, s op b.w); \
+    }
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, uchar, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, uchar, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, uchar, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, char, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, char, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, char, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, ushort, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, ushort, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, ushort, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, short, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, short, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, short, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, int, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, int, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, int, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, uint, uint, uint)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, uint, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, uint, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, float, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, float, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(+, double, double, double)
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, uchar, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, uchar, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, uchar, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, char, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, char, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, char, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, ushort, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, ushort, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, ushort, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, short, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, short, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, short, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, int, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, int, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, int, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, uint, uint, uint)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, uint, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, uint, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, float, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, float, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(-, double, double, double)
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, uchar, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, uchar, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, uchar, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, char, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, char, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, char, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, ushort, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, ushort, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, ushort, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, short, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, short, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, short, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, int, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, int, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, int, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, uint, uint, uint)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, uint, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, uint, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, float, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, float, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(*, double, double, double)
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, uchar, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, uchar, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, uchar, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, char, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, char, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, char, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, ushort, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, ushort, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, ushort, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, short, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, short, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, short, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, int, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, int, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, int, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, uint, uint, uint)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, uint, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, uint, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, float, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, float, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(/, double, double, double)
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(==, uchar, uchar, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(==, char, char, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(==, ushort, ushort, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(==, short, short, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(==, int, int, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(==, uint, uint, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(==, float, float, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(==, double, double, uchar)
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(!=, uchar, uchar, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(!=, char, char, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(!=, ushort, ushort, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(!=, short, short, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(!=, int, int, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(!=, uint, uint, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(!=, float, float, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(!=, double, double, uchar)
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(>, uchar, uchar, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(>, char, char, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(>, ushort, ushort, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(>, short, short, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(>, int, int, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(>, uint, uint, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(>, float, float, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(>, double, double, uchar)
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(<, uchar, uchar, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(<, char, char, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(<, ushort, ushort, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(<, short, short, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(<, int, int, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(<, uint, uint, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(<, float, float, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(<, double, double, uchar)
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(>=, uchar, uchar, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(>=, char, char, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(>=, ushort, ushort, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(>=, short, short, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(>=, int, int, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(>=, uint, uint, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(>=, float, float, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(>=, double, double, uchar)
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(<=, uchar, uchar, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(<=, char, char, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(<=, ushort, ushort, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(<=, short, short, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(<=, int, int, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(<=, uint, uint, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(<=, float, float, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(<=, double, double, uchar)
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(&&, uchar, uchar, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(&&, char, char, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(&&, ushort, ushort, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(&&, short, short, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(&&, int, int, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(&&, uint, uint, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(&&, float, float, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(&&, double, double, uchar)
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(||, uchar, uchar, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(||, char, char, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(||, ushort, ushort, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(||, short, short, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(||, int, int, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(||, uint, uint, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(||, float, float, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(||, double, double, uchar)
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(&, uchar, uchar, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(&, char, char, char)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(&, ushort, ushort, ushort)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(&, short, short, short)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(&, int, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(&, uint, uint, uint)
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(|, uchar, uchar, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(|, char, char, char)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(|, ushort, ushort, ushort)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(|, short, short, short)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(|, int, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(|, uint, uint, uint)
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(^, uchar, uchar, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(^, char, char, char)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(^, ushort, ushort, ushort)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(^, short, short, short)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(^, int, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP(^, uint, uint, uint)
+
+#undef CV_CUDEV_IMPLEMENT_SCALAR_BINARY_OP
+
+// binary function (vec & vec)
+
+#define CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(func_name, func, input_type, output_type) \
+    __device__ __forceinline__ output_type ## 1 func_name(const input_type ## 1 & a, const input_type ## 1 & b) \
+    { \
+        return VecTraits<output_type ## 1>::make(func (a.x, b.x)); \
+    } \
+    __device__ __forceinline__ output_type ## 2 func_name(const input_type ## 2 & a, const input_type ## 2 & b) \
+    { \
+        return VecTraits<output_type ## 2>::make(func (a.x, b.x), func (a.y, b.y)); \
+    } \
+    __device__ __forceinline__ output_type ## 3 func_name(const input_type ## 3 & a, const input_type ## 3 & b) \
+    { \
+        return VecTraits<output_type ## 3>::make(func (a.x, b.x), func (a.y, b.y), func (a.z, b.z)); \
+    } \
+    __device__ __forceinline__ output_type ## 4 func_name(const input_type ## 4 & a, const input_type ## 4 & b) \
+    { \
+        return VecTraits<output_type ## 4>::make(func (a.x, b.x), func (a.y, b.y), func (a.z, b.z), func (a.w, b.w)); \
+    }
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(max, ::max, uchar, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(max, ::max, char, char)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(max, ::max, ushort, ushort)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(max, ::max, short, short)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(max, ::max, uint, uint)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(max, ::max, int, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(max, ::fmaxf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(max, ::fmax, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(min, ::min, uchar, uchar)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(min, ::min, char, char)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(min, ::min, ushort, ushort)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(min, ::min, short, short)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(min, ::min, uint, uint)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(min, ::min, int, int)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(min, ::fminf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(min, ::fmin, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(hypot, ::hypotf, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(hypot, ::hypotf, char, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(hypot, ::hypotf, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(hypot, ::hypotf, short, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(hypot, ::hypotf, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(hypot, ::hypotf, int, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(hypot, ::hypotf, float, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(hypot, ::hypot, double, double)
+
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(atan2, ::atan2f, uchar, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(atan2, ::atan2f, char, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(atan2, ::atan2f, ushort, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(atan2, ::atan2f, short, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(atan2, ::atan2f, uint, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(atan2, ::atan2f, int, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(atan2, ::atan2f, float, float)
+CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC(atan2, ::atan2, double, double)
+
+#undef CV_CUDEV_IMPLEMENT_VEC_BINARY_FUNC
+
+// binary function (vec & scalar)
+
+#define CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(func_name, func, input_type, scalar_type, output_type) \
+    __device__ __forceinline__ output_type ## 1 func_name(const input_type ## 1 & a, scalar_type s) \
+    { \
+        return VecTraits<output_type ## 1>::make(func ((output_type) a.x, (output_type) s)); \
+    } \
+    __device__ __forceinline__ output_type ## 1 func_name(scalar_type s, const input_type ## 1 & b) \
+    { \
+        return VecTraits<output_type ## 1>::make(func ((output_type) s, (output_type) b.x)); \
+    } \
+    __device__ __forceinline__ output_type ## 2 func_name(const input_type ## 2 & a, scalar_type s) \
+    { \
+        return VecTraits<output_type ## 2>::make(func ((output_type) a.x, (output_type) s), func ((output_type) a.y, (output_type) s)); \
+    } \
+    __device__ __forceinline__ output_type ## 2 func_name(scalar_type s, const input_type ## 2 & b) \
+    { \
+        return VecTraits<output_type ## 2>::make(func ((output_type) s, (output_type) b.x), func ((output_type) s, (output_type) b.y)); \
+    } \
+    __device__ __forceinline__ output_type ## 3 func_name(const input_type ## 3 & a, scalar_type s) \
+    { \
+        return VecTraits<output_type ## 3>::make(func ((output_type) a.x, (output_type) s), func ((output_type) a.y, (output_type) s), func ((output_type) a.z, (output_type) s)); \
+    } \
+    __device__ __forceinline__ output_type ## 3 func_name(scalar_type s, const input_type ## 3 & b) \
+    { \
+        return VecTraits<output_type ## 3>::make(func ((output_type) s, (output_type) b.x), func ((output_type) s, (output_type) b.y), func ((output_type) s, (output_type) b.z)); \
+    } \
+    __device__ __forceinline__ output_type ## 4 func_name(const input_type ## 4 & a, scalar_type s) \
+    { \
+        return VecTraits<output_type ## 4>::make(func ((output_type) a.x, (output_type) s), func ((output_type) a.y, (output_type) s), func ((output_type) a.z, (output_type) s), func ((output_type) a.w, (output_type) s)); \
+    } \
+    __device__ __forceinline__ output_type ## 4 func_name(scalar_type s, const input_type ## 4 & b) \
+    { \
+        return VecTraits<output_type ## 4>::make(func ((output_type) s, (output_type) b.x), func ((output_type) s, (output_type) b.y), func ((output_type) s, (output_type) b.z), func ((output_type) s, (output_type) b.w)); \
+    }
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::max, uchar, uchar, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::fmaxf, uchar, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::fmax, uchar, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::max, char, char, char)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::fmaxf, char, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::fmax, char, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::max, ushort, ushort, ushort)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::fmaxf, ushort, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::fmax, ushort, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::max, short, short, short)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::fmaxf, short, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::fmax, short, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::max, uint, uint, uint)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::fmaxf, uint, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::fmax, uint, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::max, int, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::fmaxf, int, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::fmax, int, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::fmaxf, float, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::fmax, float, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(max, ::fmax, double, double, double)
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::min, uchar, uchar, uchar)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::fminf, uchar, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::fmin, uchar, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::min, char, char, char)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::fminf, char, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::fmin, char, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::min, ushort, ushort, ushort)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::fminf, ushort, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::fmin, ushort, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::min, short, short, short)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::fminf, short, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::fmin, short, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::min, uint, uint, uint)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::fminf, uint, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::fmin, uint, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::min, int, int, int)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::fminf, int, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::fmin, int, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::fminf, float, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::fmin, float, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(min, ::fmin, double, double, double)
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(hypot, ::hypotf, uchar, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(hypot, ::hypot, uchar, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(hypot, ::hypotf, char, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(hypot, ::hypot, char, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(hypot, ::hypotf, ushort, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(hypot, ::hypot, ushort, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(hypot, ::hypotf, short, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(hypot, ::hypot, short, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(hypot, ::hypotf, uint, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(hypot, ::hypot, uint, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(hypot, ::hypotf, int, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(hypot, ::hypot, int, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(hypot, ::hypotf, float, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(hypot, ::hypot, float, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(hypot, ::hypot, double, double, double)
+
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(atan2, ::atan2f, uchar, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(atan2, ::atan2, uchar, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(atan2, ::atan2f, char, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(atan2, ::atan2, char, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(atan2, ::atan2f, ushort, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(atan2, ::atan2, ushort, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(atan2, ::atan2f, short, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(atan2, ::atan2, short, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(atan2, ::atan2f, uint, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(atan2, ::atan2, uint, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(atan2, ::atan2f, int, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(atan2, ::atan2, int, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(atan2, ::atan2f, float, float, float)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(atan2, ::atan2, float, double, double)
+CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC(atan2, ::atan2, double, double, double)
+
+#undef CV_CUDEV_IMPLEMENT_SCALAR_BINARY_FUNC
+
+}}} // namespace cv { namespace cuda { namespace device
+
+//! @endcond
+
+#endif // OPENCV_CUDA_VECMATH_HPP

model_zoo/official/lite/app/src/main/cpp/include/opencv2/core/cuda/vec_traits.hpp

@@ -0,0 +1,288 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CUDA_VEC_TRAITS_HPP
+#define OPENCV_CUDA_VEC_TRAITS_HPP
+
+#include "common.hpp"
+
+/** @file
+ * @deprecated Use @ref cudev instead.
+ */
+
+//! @cond IGNORED
+
+namespace cv { namespace cuda { namespace device
+{
+    template<typename T, int N> struct TypeVec;
+
+    struct __align__(8) uchar8
+    {
+        uchar a0, a1, a2, a3, a4, a5, a6, a7;
+    };
+    static __host__ __device__ __forceinline__ uchar8 make_uchar8(uchar a0, uchar a1, uchar a2, uchar a3, uchar a4, uchar a5, uchar a6, uchar a7)
+    {
+        uchar8 val = {a0, a1, a2, a3, a4, a5, a6, a7};
+        return val;
+    }
+    struct __align__(8) char8
+    {
+        schar a0, a1, a2, a3, a4, a5, a6, a7;
+    };
+    static __host__ __device__ __forceinline__ char8 make_char8(schar a0, schar a1, schar a2, schar a3, schar a4, schar a5, schar a6, schar a7)
+    {
+        char8 val = {a0, a1, a2, a3, a4, a5, a6, a7};
+        return val;
+    }
+    struct __align__(16) ushort8
+    {
+        ushort a0, a1, a2, a3, a4, a5, a6, a7;
+    };
+    static __host__ __device__ __forceinline__ ushort8 make_ushort8(ushort a0, ushort a1, ushort a2, ushort a3, ushort a4, ushort a5, ushort a6, ushort a7)
+    {
+        ushort8 val = {a0, a1, a2, a3, a4, a5, a6, a7};
+        return val;
+    }
+    struct __align__(16) short8
+    {
+        short a0, a1, a2, a3, a4, a5, a6, a7;
+    };
+    static __host__ __device__ __forceinline__ short8 make_short8(short a0, short a1, short a2, short a3, short a4, short a5, short a6, short a7)
+    {
+        short8 val = {a0, a1, a2, a3, a4, a5, a6, a7};
+        return val;
+    }
+    struct __align__(32) uint8
+    {
+        uint a0, a1, a2, a3, a4, a5, a6, a7;
+    };
+    static __host__ __device__ __forceinline__ uint8 make_uint8(uint a0, uint a1, uint a2, uint a3, uint a4, uint a5, uint a6, uint a7)
+    {
+        uint8 val = {a0, a1, a2, a3, a4, a5, a6, a7};
+        return val;
+    }
+    struct __align__(32) int8
+    {
+        int a0, a1, a2, a3, a4, a5, a6, a7;
+    };
+    static __host__ __device__ __forceinline__ int8 make_int8(int a0, int a1, int a2, int a3, int a4, int a5, int a6, int a7)
+    {
+        int8 val = {a0, a1, a2, a3, a4, a5, a6, a7};
+        return val;
+    }
+    struct __align__(32) float8
+    {
+        float a0, a1, a2, a3, a4, a5, a6, a7;
+    };
+    static __host__ __device__ __forceinline__ float8 make_float8(float a0, float a1, float a2, float a3, float a4, float a5, float a6, float a7)
+    {
+        float8 val = {a0, a1, a2, a3, a4, a5, a6, a7};
+        return val;
+    }
+    struct double8
+    {
+        double a0, a1, a2, a3, a4, a5, a6, a7;
+    };
+    static __host__ __device__ __forceinline__ double8 make_double8(double a0, double a1, double a2, double a3, double a4, double a5, double a6, double a7)
+    {
+        double8 val = {a0, a1, a2, a3, a4, a5, a6, a7};
+        return val;
+    }
+
+#define OPENCV_CUDA_IMPLEMENT_TYPE_VEC(type) \
+    template<> struct TypeVec<type, 1> { typedef type vec_type; }; \
+    template<> struct TypeVec<type ## 1, 1> { typedef type ## 1 vec_type; }; \
+    template<> struct TypeVec<type, 2> { typedef type ## 2 vec_type; }; \
+    template<> struct TypeVec<type ## 2, 2> { typedef type ## 2 vec_type; }; \
+    template<> struct TypeVec<type, 3> { typedef type ## 3 vec_type; }; \
+    template<> struct TypeVec<type ## 3, 3> { typedef type ## 3 vec_type; }; \
+    template<> struct TypeVec<type, 4> { typedef type ## 4 vec_type; }; \
+    template<> struct TypeVec<type ## 4, 4> { typedef type ## 4 vec_type; }; \
+    template<> struct TypeVec<type, 8> { typedef type ## 8 vec_type; }; \
+    template<> struct TypeVec<type ## 8, 8> { typedef type ## 8 vec_type; };
+
+    OPENCV_CUDA_IMPLEMENT_TYPE_VEC(uchar)
+    OPENCV_CUDA_IMPLEMENT_TYPE_VEC(char)
+    OPENCV_CUDA_IMPLEMENT_TYPE_VEC(ushort)
+    OPENCV_CUDA_IMPLEMENT_TYPE_VEC(short)
+    OPENCV_CUDA_IMPLEMENT_TYPE_VEC(int)
+    OPENCV_CUDA_IMPLEMENT_TYPE_VEC(uint)
+    OPENCV_CUDA_IMPLEMENT_TYPE_VEC(float)
+    OPENCV_CUDA_IMPLEMENT_TYPE_VEC(double)
+
+    #undef OPENCV_CUDA_IMPLEMENT_TYPE_VEC
+
+    template<> struct TypeVec<schar, 1> { typedef schar vec_type; };
+    template<> struct TypeVec<schar, 2> { typedef char2 vec_type; };
+    template<> struct TypeVec<schar, 3> { typedef char3 vec_type; };
+    template<> struct TypeVec<schar, 4> { typedef char4 vec_type; };
+    template<> struct TypeVec<schar, 8> { typedef char8 vec_type; };
+
+    template<> struct TypeVec<bool, 1> { typedef uchar vec_type; };
+    template<> struct TypeVec<bool, 2> { typedef uchar2 vec_type; };
+    template<> struct TypeVec<bool, 3> { typedef uchar3 vec_type; };
+    template<> struct TypeVec<bool, 4> { typedef uchar4 vec_type; };
+    template<> struct TypeVec<bool, 8> { typedef uchar8 vec_type; };
+
+    template<typename T> struct VecTraits;
+
+#define OPENCV_CUDA_IMPLEMENT_VEC_TRAITS(type) \
+    template<> struct VecTraits<type> \
+    { \
+        typedef type elem_type; \
+        enum {cn=1}; \
+        static __device__ __host__ __forceinline__ type all(type v) {return v;} \
+        static __device__ __host__ __forceinline__ type make(type x) {return x;} \
+        static __device__ __host__ __forceinline__ type make(const type* v) {return *v;} \
+    }; \
+    template<> struct VecTraits<type ## 1> \
+    { \
+        typedef type elem_type; \
+        enum {cn=1}; \
+        static __device__ __host__ __forceinline__ type ## 1 all(type v) {return make_ ## type ## 1(v);} \
+        static __device__ __host__ __forceinline__ type ## 1 make(type x) {return make_ ## type ## 1(x);} \
+        static __device__ __host__ __forceinline__ type ## 1 make(const type* v) {return make_ ## type ## 1(*v);} \
+    }; \
+    template<> struct VecTraits<type ## 2> \
+    { \
+        typedef type elem_type; \
+        enum {cn=2}; \
+        static __device__ __host__ __forceinline__ type ## 2 all(type v) {return make_ ## type ## 2(v, v);} \
+        static __device__ __host__ __forceinline__ type ## 2 make(type x, type y) {return make_ ## type ## 2(x, y);} \
+        static __device__ __host__ __forceinline__ type ## 2 make(const type* v) {return make_ ## type ## 2(v[0], v[1]);} \
+    }; \
+    template<> struct VecTraits<type ## 3> \
+    { \
+        typedef type elem_type; \
+        enum {cn=3}; \
+        static __device__ __host__ __forceinline__ type ## 3 all(type v) {return make_ ## type ## 3(v, v, v);} \
+        static __device__ __host__ __forceinline__ type ## 3 make(type x, type y, type z) {return make_ ## type ## 3(x, y, z);} \
+        static __device__ __host__ __forceinline__ type ## 3 make(const type* v) {return make_ ## type ## 3(v[0], v[1], v[2]);} \
+    }; \
+    template<> struct VecTraits<type ## 4> \
+    { \
+        typedef type elem_type; \
+        enum {cn=4}; \
+        static __device__ __host__ __forceinline__ type ## 4 all(type v) {return make_ ## type ## 4(v, v, v, v);} \
+        static __device__ __host__ __forceinline__ type ## 4 make(type x, type y, type z, type w) {return make_ ## type ## 4(x, y, z, w);} \
+        static __device__ __host__ __forceinline__ type ## 4 make(const type* v) {return make_ ## type ## 4(v[0], v[1], v[2], v[3]);} \
+    }; \
+    template<> struct VecTraits<type ## 8> \
+    { \
+        typedef type elem_type; \
+        enum {cn=8}; \
+        static __device__ __host__ __forceinline__ type ## 8 all(type v) {return make_ ## type ## 8(v, v, v, v, v, v, v, v);} \
+        static __device__ __host__ __forceinline__ type ## 8 make(type a0, type a1, type a2, type a3, type a4, type a5, type a6, type a7) {return make_ ## type ## 8(a0, a1, a2, a3, a4, a5, a6, a7);} \
+        static __device__ __host__ __forceinline__ type ## 8 make(const type* v) {return make_ ## type ## 8(v[0], v[1], v[2], v[3], v[4], v[5], v[6], v[7]);} \
+    };
+
+    OPENCV_CUDA_IMPLEMENT_VEC_TRAITS(uchar)
+    OPENCV_CUDA_IMPLEMENT_VEC_TRAITS(ushort)
+    OPENCV_CUDA_IMPLEMENT_VEC_TRAITS(short)
+    OPENCV_CUDA_IMPLEMENT_VEC_TRAITS(int)
+    OPENCV_CUDA_IMPLEMENT_VEC_TRAITS(uint)
+    OPENCV_CUDA_IMPLEMENT_VEC_TRAITS(float)
+    OPENCV_CUDA_IMPLEMENT_VEC_TRAITS(double)
+
+    #undef OPENCV_CUDA_IMPLEMENT_VEC_TRAITS
+
+    template<> struct VecTraits<char>
+    {
+        typedef char elem_type;
+        enum {cn=1};
+        static __device__ __host__ __forceinline__ char all(char v) {return v;}
+        static __device__ __host__ __forceinline__ char make(char x) {return x;}
+        static __device__ __host__ __forceinline__ char make(const char* x) {return *x;}
+    };
+    template<> struct VecTraits<schar>
+    {
+        typedef schar elem_type;
+        enum {cn=1};
+        static __device__ __host__ __forceinline__ schar all(schar v) {return v;}
+        static __device__ __host__ __forceinline__ schar make(schar x) {return x;}
+        static __device__ __host__ __forceinline__ schar make(const schar* x) {return *x;}
+    };
+    template<> struct VecTraits<char1>
+    {
+        typedef schar elem_type;
+        enum {cn=1};
+        static __device__ __host__ __forceinline__ char1 all(schar v) {return make_char1(v);}
+        static __device__ __host__ __forceinline__ char1 make(schar x) {return make_char1(x);}
+        static __device__ __host__ __forceinline__ char1 make(const schar* v) {return make_char1(v[0]);}
+    };
+    template<> struct VecTraits<char2>
+    {
+        typedef schar elem_type;
+        enum {cn=2};
+        static __device__ __host__ __forceinline__ char2 all(schar v) {return make_char2(v, v);}
+        static __device__ __host__ __forceinline__ char2 make(schar x, schar y) {return make_char2(x, y);}
+        static __device__ __host__ __forceinline__ char2 make(const schar* v) {return make_char2(v[0], v[1]);}
+    };
+    template<> struct VecTraits<char3>
+    {
+        typedef schar elem_type;
+        enum {cn=3};
+        static __device__ __host__ __forceinline__ char3 all(schar v) {return make_char3(v, v, v);}
+        static __device__ __host__ __forceinline__ char3 make(schar x, schar y, schar z) {return make_char3(x, y, z);}
+        static __device__ __host__ __forceinline__ char3 make(const schar* v) {return make_char3(v[0], v[1], v[2]);}
+    };
+    template<> struct VecTraits<char4>
+    {
+        typedef schar elem_type;
+        enum {cn=4};
+        static __device__ __host__ __forceinline__ char4 all(schar v) {return make_char4(v, v, v, v);}
+        static __device__ __host__ __forceinline__ char4 make(schar x, schar y, schar z, schar w) {return make_char4(x, y, z, w);}
+        static __device__ __host__ __forceinline__ char4 make(const schar* v) {return make_char4(v[0], v[1], v[2], v[3]);}
+    };
+    template<> struct VecTraits<char8>
+    {
+        typedef schar elem_type;
+        enum {cn=8};
+        static __device__ __host__ __forceinline__ char8 all(schar v) {return make_char8(v, v, v, v, v, v, v, v);}
+        static __device__ __host__ __forceinline__ char8 make(schar a0, schar a1, schar a2, schar a3, schar a4, schar a5, schar a6, schar a7) {return make_char8(a0, a1, a2, a3, a4, a5, a6, a7);}
+        static __device__ __host__ __forceinline__ char8 make(const schar* v) {return make_char8(v[0], v[1], v[2], v[3], v[4], v[5], v[6], v[7]);}
+    };
+}}} // namespace cv { namespace cuda { namespace cudev
+
+//! @endcond
+
+#endif // OPENCV_CUDA_VEC_TRAITS_HPP

model_zoo/official/lite/app/src/main/cpp/include/opencv2/core/cuda/warp.hpp

@@ -0,0 +1,139 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CUDA_DEVICE_WARP_HPP
+#define OPENCV_CUDA_DEVICE_WARP_HPP
+
+/** @file
+ * @deprecated Use @ref cudev instead.
+ */
+
+//! @cond IGNORED
+
+namespace cv { namespace cuda { namespace device
+{
+    struct Warp
+    {
+        enum
+        {
+            LOG_WARP_SIZE = 5,
+            WARP_SIZE     = 1 << LOG_WARP_SIZE,
+            STRIDE        = WARP_SIZE
+        };
+
+        /** \brief Returns the warp lane ID of the calling thread. */
+        static __device__ __forceinline__ unsigned int laneId()
+        {
+            unsigned int ret;
+            asm("mov.u32 %0, %%laneid;" : "=r"(ret) );
+            return ret;
+        }
+
+        template<typename It, typename T>
+        static __device__ __forceinline__ void fill(It beg, It end, const T& value)
+        {
+            for(It t = beg + laneId(); t < end; t += STRIDE)
+                *t = value;
+        }
+
+        template<typename InIt, typename OutIt>
+        static __device__ __forceinline__ OutIt copy(InIt beg, InIt end, OutIt out)
+        {
+            for(InIt t = beg + laneId(); t < end; t += STRIDE, out += STRIDE)
+                *out = *t;
+            return out;
+        }
+
+        template<typename InIt, typename OutIt, class UnOp>
+        static __device__ __forceinline__ OutIt transform(InIt beg, InIt end, OutIt out, UnOp op)
+        {
+            for(InIt t = beg + laneId(); t < end; t += STRIDE, out += STRIDE)
+                *out = op(*t);
+            return out;
+        }
+
+        template<typename InIt1, typename InIt2, typename OutIt, class BinOp>
+        static __device__ __forceinline__ OutIt transform(InIt1 beg1, InIt1 end1, InIt2 beg2, OutIt out, BinOp op)
+        {
+            unsigned int lane = laneId();
+
+            InIt1 t1 = beg1 + lane;
+            InIt2 t2 = beg2 + lane;
+            for(; t1 < end1; t1 += STRIDE, t2 += STRIDE, out += STRIDE)
+                *out = op(*t1, *t2);
+            return out;
+        }
+
+        template <class T, class BinOp>
+        static __device__ __forceinline__ T reduce(volatile T *ptr, BinOp op)
+        {
+            const unsigned int lane = laneId();
+
+            if (lane < 16)
+            {
+                T partial = ptr[lane];
+
+                ptr[lane] = partial = op(partial, ptr[lane + 16]);
+                ptr[lane] = partial = op(partial, ptr[lane + 8]);
+                ptr[lane] = partial = op(partial, ptr[lane + 4]);
+                ptr[lane] = partial = op(partial, ptr[lane + 2]);
+                ptr[lane] = partial = op(partial, ptr[lane + 1]);
+            }
+
+            return *ptr;
+        }
+
+        template<typename OutIt, typename T>
+        static __device__ __forceinline__ void yota(OutIt beg, OutIt end, T value)
+        {
+            unsigned int lane = laneId();
+            value += lane;
+
+            for(OutIt t = beg + lane; t < end; t += STRIDE, value += STRIDE)
+                *t = value;
+        }
+    };
+}}} // namespace cv { namespace cuda { namespace cudev
+
+//! @endcond
+
+#endif /* OPENCV_CUDA_DEVICE_WARP_HPP */

model_zoo/official/lite/app/src/main/cpp/include/opencv2/core/cuda/warp_reduce.hpp

@@ -0,0 +1,76 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CUDA_WARP_REDUCE_HPP__
+#define OPENCV_CUDA_WARP_REDUCE_HPP__
+
+/** @file
+ * @deprecated Use @ref cudev instead.
+ */
+
+//! @cond IGNORED
+
+namespace cv { namespace cuda { namespace device
+{
+    template <class T>
+    __device__ __forceinline__ T warp_reduce(volatile T *ptr , const unsigned int tid = threadIdx.x)
+    {
+        const unsigned int lane = tid & 31; // index of thread in warp (0..31)
+
+        if (lane < 16)
+        {
+            T partial = ptr[tid];
+
+            ptr[tid] = partial = partial + ptr[tid + 16];
+            ptr[tid] = partial = partial + ptr[tid + 8];
+            ptr[tid] = partial = partial + ptr[tid + 4];
+            ptr[tid] = partial = partial + ptr[tid + 2];
+            ptr[tid] = partial = partial + ptr[tid + 1];
+        }
+
+        return ptr[tid - lane];
+    }
+}}} // namespace cv { namespace cuda { namespace cudev {
+
+//! @endcond
+
+#endif /* OPENCV_CUDA_WARP_REDUCE_HPP__ */

model_zoo/official/lite/app/src/main/cpp/include/opencv2/core/cuda/warp_shuffle.hpp

@@ -0,0 +1,162 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CUDA_WARP_SHUFFLE_HPP
+#define OPENCV_CUDA_WARP_SHUFFLE_HPP
+
+/** @file
+ * @deprecated Use @ref cudev instead.
+ */
+
+//! @cond IGNORED
+
+namespace cv { namespace cuda { namespace device
+{
+#if __CUDACC_VER_MAJOR__ >= 9
+#  define __shfl(x, y, z) __shfl_sync(0xFFFFFFFFU, x, y, z)
+#  define __shfl_up(x, y, z) __shfl_up_sync(0xFFFFFFFFU, x, y, z)
+#  define __shfl_down(x, y, z) __shfl_down_sync(0xFFFFFFFFU, x, y, z)
+#endif
+    template <typename T>
+    __device__ __forceinline__ T shfl(T val, int srcLane, int width = warpSize)
+    {
+    #if defined __CUDA_ARCH__ && __CUDA_ARCH__ >= 300
+        return __shfl(val, srcLane, width);
+    #else
+        return T();
+    #endif
+    }
+    __device__ __forceinline__ unsigned int shfl(unsigned int val, int srcLane, int width = warpSize)
+    {
+    #if defined __CUDA_ARCH__ && __CUDA_ARCH__ >= 300
+        return (unsigned int) __shfl((int) val, srcLane, width);
+    #else
+        return 0;
+    #endif
+    }
+    __device__ __forceinline__ double shfl(double val, int srcLane, int width = warpSize)
+    {
+    #if defined __CUDA_ARCH__ && __CUDA_ARCH__ >= 300
+        int lo = __double2loint(val);
+        int hi = __double2hiint(val);
+
+        lo = __shfl(lo, srcLane, width);
+        hi = __shfl(hi, srcLane, width);
+
+        return __hiloint2double(hi, lo);
+    #else
+        return 0.0;
+    #endif
+    }
+
+    template <typename T>
+    __device__ __forceinline__ T shfl_down(T val, unsigned int delta, int width = warpSize)
+    {
+    #if defined __CUDA_ARCH__ && __CUDA_ARCH__ >= 300
+        return __shfl_down(val, delta, width);
+    #else
+        return T();
+    #endif
+    }
+    __device__ __forceinline__ unsigned int shfl_down(unsigned int val, unsigned int delta, int width = warpSize)
+    {
+    #if defined __CUDA_ARCH__ && __CUDA_ARCH__ >= 300
+        return (unsigned int) __shfl_down((int) val, delta, width);
+    #else
+        return 0;
+    #endif
+    }
+    __device__ __forceinline__ double shfl_down(double val, unsigned int delta, int width = warpSize)
+    {
+    #if defined __CUDA_ARCH__ && __CUDA_ARCH__ >= 300
+        int lo = __double2loint(val);
+        int hi = __double2hiint(val);
+
+        lo = __shfl_down(lo, delta, width);
+        hi = __shfl_down(hi, delta, width);
+
+        return __hiloint2double(hi, lo);
+    #else
+        return 0.0;
+    #endif
+    }
+
+    template <typename T>
+    __device__ __forceinline__ T shfl_up(T val, unsigned int delta, int width = warpSize)
+    {
+    #if defined __CUDA_ARCH__ && __CUDA_ARCH__ >= 300
+        return __shfl_up(val, delta, width);
+    #else
+        return T();
+    #endif
+    }
+    __device__ __forceinline__ unsigned int shfl_up(unsigned int val, unsigned int delta, int width = warpSize)
+    {
+    #if defined __CUDA_ARCH__ && __CUDA_ARCH__ >= 300
+        return (unsigned int) __shfl_up((int) val, delta, width);
+    #else
+        return 0;
+    #endif
+    }
+    __device__ __forceinline__ double shfl_up(double val, unsigned int delta, int width = warpSize)
+    {
+    #if defined __CUDA_ARCH__ && __CUDA_ARCH__ >= 300
+        int lo = __double2loint(val);
+        int hi = __double2hiint(val);
+
+        lo = __shfl_up(lo, delta, width);
+        hi = __shfl_up(hi, delta, width);
+
+        return __hiloint2double(hi, lo);
+    #else
+        return 0.0;
+    #endif
+    }
+}}}
+
+#  undef __shfl
+#  undef __shfl_up
+#  undef __shfl_down
+
+//! @endcond
+
+#endif // OPENCV_CUDA_WARP_SHUFFLE_HPP

model_zoo/official/lite/app/src/main/cpp/include/opencv2/core/cuda_stream_accessor.hpp

@@ -0,0 +1,86 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CORE_CUDA_STREAM_ACCESSOR_HPP
+#define OPENCV_CORE_CUDA_STREAM_ACCESSOR_HPP
+
+#ifndef __cplusplus
+#  error cuda_stream_accessor.hpp header must be compiled as C++
+#endif
+
+/** @file cuda_stream_accessor.hpp
+ * This is only header file that depends on CUDA Runtime API. All other headers are independent.
+ */
+
+#include <cuda_runtime.h>
+#include "opencv2/core/cuda.hpp"
+
+namespace cv
+{
+    namespace cuda
+    {
+
+//! @addtogroup cudacore_struct
+//! @{
+
+        /** @brief Class that enables getting cudaStream_t from cuda::Stream
+         */
+        struct StreamAccessor
+        {
+            CV_EXPORTS static cudaStream_t getStream(const Stream& stream);
+            CV_EXPORTS static Stream wrapStream(cudaStream_t stream);
+        };
+
+        /** @brief Class that enables getting cudaEvent_t from cuda::Event
+         */
+        struct EventAccessor
+        {
+            CV_EXPORTS static cudaEvent_t getEvent(const Event& event);
+            CV_EXPORTS static Event wrapEvent(cudaEvent_t event);
+        };
+
+//! @}
+
+    }
+}
+
+#endif /* OPENCV_CORE_CUDA_STREAM_ACCESSOR_HPP */

model_zoo/official/lite/app/src/main/cpp/include/opencv2/core/cuda_types.hpp

@@ -0,0 +1,144 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CORE_CUDA_TYPES_HPP
+#define OPENCV_CORE_CUDA_TYPES_HPP
+
+#ifndef __cplusplus
+#  error cuda_types.hpp header must be compiled as C++
+#endif
+
+#if defined(__OPENCV_BUILD) && defined(__clang__)
+#pragma clang diagnostic ignored "-Winconsistent-missing-override"
+#endif
+#if defined(__OPENCV_BUILD) && defined(__GNUC__) && __GNUC__ >= 5
+#pragma GCC diagnostic ignored "-Wsuggest-override"
+#endif
+
+/** @file
+ * @deprecated Use @ref cudev instead.
+ */
+
+//! @cond IGNORED
+
+#ifdef __CUDACC__
+    #define __CV_CUDA_HOST_DEVICE__ __host__ __device__ __forceinline__
+#else
+    #define __CV_CUDA_HOST_DEVICE__
+#endif
+
+namespace cv
+{
+    namespace cuda
+    {
+
+        // Simple lightweight structures that encapsulates information about an image on device.
+        // It is intended to pass to nvcc-compiled code. GpuMat depends on headers that nvcc can't compile
+
+        template <typename T> struct DevPtr
+        {
+            typedef T elem_type;
+            typedef int index_type;
+
+            enum { elem_size = sizeof(elem_type) };
+
+            T* data;
+
+            __CV_CUDA_HOST_DEVICE__ DevPtr() : data(0) {}
+            __CV_CUDA_HOST_DEVICE__ DevPtr(T* data_) : data(data_) {}
+
+            __CV_CUDA_HOST_DEVICE__ size_t elemSize() const { return elem_size; }
+            __CV_CUDA_HOST_DEVICE__ operator       T*()       { return data; }
+            __CV_CUDA_HOST_DEVICE__ operator const T*() const { return data; }
+        };
+
+        template <typename T> struct PtrSz : public DevPtr<T>
+        {
+            __CV_CUDA_HOST_DEVICE__ PtrSz() : size(0) {}
+            __CV_CUDA_HOST_DEVICE__ PtrSz(T* data_, size_t size_) : DevPtr<T>(data_), size(size_) {}
+
+            size_t size;
+        };
+
+        template <typename T> struct PtrStep : public DevPtr<T>
+        {
+            __CV_CUDA_HOST_DEVICE__ PtrStep() : step(0) {}
+            __CV_CUDA_HOST_DEVICE__ PtrStep(T* data_, size_t step_) : DevPtr<T>(data_), step(step_) {}
+
+            size_t step;
+
+            __CV_CUDA_HOST_DEVICE__       T* ptr(int y = 0)       { return (      T*)( (      char*)(((DevPtr<T>*)this)->data) + y * step); }
+            __CV_CUDA_HOST_DEVICE__ const T* ptr(int y = 0) const { return (const T*)( (const char*)(((DevPtr<T>*)this)->data) + y * step); }
+
+            __CV_CUDA_HOST_DEVICE__       T& operator ()(int y, int x)       { return ptr(y)[x]; }
+            __CV_CUDA_HOST_DEVICE__ const T& operator ()(int y, int x) const { return ptr(y)[x]; }
+        };
+
+        template <typename T> struct PtrStepSz : public PtrStep<T>
+        {
+            __CV_CUDA_HOST_DEVICE__ PtrStepSz() : cols(0), rows(0) {}
+            __CV_CUDA_HOST_DEVICE__ PtrStepSz(int rows_, int cols_, T* data_, size_t step_)
+                : PtrStep<T>(data_, step_), cols(cols_), rows(rows_) {}
+
+            template <typename U>
+            explicit PtrStepSz(const PtrStepSz<U>& d) : PtrStep<T>((T*)d.data, d.step), cols(d.cols), rows(d.rows){}
+
+            int cols;
+            int rows;
+        };
+
+        typedef PtrStepSz<unsigned char> PtrStepSzb;
+        typedef PtrStepSz<unsigned short> PtrStepSzus;
+        typedef PtrStepSz<float> PtrStepSzf;
+        typedef PtrStepSz<int> PtrStepSzi;
+
+        typedef PtrStep<unsigned char> PtrStepb;
+        typedef PtrStep<unsigned short> PtrStepus;
+        typedef PtrStep<float> PtrStepf;
+        typedef PtrStep<int> PtrStepi;
+
+    }
+}
+
+//! @endcond
+
+#endif /* OPENCV_CORE_CUDA_TYPES_HPP */

model_zoo/official/lite/app/src/main/cpp/include/opencv2/core/cv_cpu_dispatch.h

@@ -0,0 +1,345 @@
+// This file is part of OpenCV project.
+// It is subject to the license terms in the LICENSE file found in the top-level directory
+// of this distribution and at http://opencv.org/license.html.
+
+#if defined __OPENCV_BUILD \
+
+#include "cv_cpu_config.h"
+#include "cv_cpu_helper.h"
+
+#ifdef CV_CPU_DISPATCH_MODE
+#define CV_CPU_OPTIMIZATION_NAMESPACE __CV_CAT(opt_, CV_CPU_DISPATCH_MODE)
+#define CV_CPU_OPTIMIZATION_NAMESPACE_BEGIN namespace __CV_CAT(opt_, CV_CPU_DISPATCH_MODE) {
+#define CV_CPU_OPTIMIZATION_NAMESPACE_END }
+#else
+#define CV_CPU_OPTIMIZATION_NAMESPACE cpu_baseline
+#define CV_CPU_OPTIMIZATION_NAMESPACE_BEGIN namespace cpu_baseline {
+#define CV_CPU_OPTIMIZATION_NAMESPACE_END }
+#define CV_CPU_BASELINE_MODE 1
+#endif
+
+
+#define __CV_CPU_DISPATCH_CHAIN_END(fn, args, mode, ...)  /* done */
+#define __CV_CPU_DISPATCH(fn, args, mode, ...) __CV_EXPAND(__CV_CPU_DISPATCH_CHAIN_ ## mode(fn, args, __VA_ARGS__))
+#define __CV_CPU_DISPATCH_EXPAND(fn, args, ...) __CV_EXPAND(__CV_CPU_DISPATCH(fn, args, __VA_ARGS__))
+#define CV_CPU_DISPATCH(fn, args, ...) __CV_CPU_DISPATCH_EXPAND(fn, args, __VA_ARGS__, END) // expand macros
+
+
+#if defined CV_ENABLE_INTRINSICS \
+    && !defined CV_DISABLE_OPTIMIZATION \
+    && !defined __CUDACC__ /* do not include SSE/AVX/NEON headers for NVCC compiler */ \
+
+#ifdef CV_CPU_COMPILE_SSE2
+#  include <emmintrin.h>
+#  define CV_MMX 1
+#  define CV_SSE 1
+#  define CV_SSE2 1
+#endif
+#ifdef CV_CPU_COMPILE_SSE3
+#  include <pmmintrin.h>
+#  define CV_SSE3 1
+#endif
+#ifdef CV_CPU_COMPILE_SSSE3
+#  include <tmmintrin.h>
+#  define CV_SSSE3 1
+#endif
+#ifdef CV_CPU_COMPILE_SSE4_1
+#  include <smmintrin.h>
+#  define CV_SSE4_1 1
+#endif
+#ifdef CV_CPU_COMPILE_SSE4_2
+#  include <nmmintrin.h>
+#  define CV_SSE4_2 1
+#endif
+#ifdef CV_CPU_COMPILE_POPCNT
+#  ifdef _MSC_VER
+#    include <nmmintrin.h>
+#    if defined(_M_X64)
+#      define CV_POPCNT_U64 _mm_popcnt_u64
+#    endif
+#    define CV_POPCNT_U32 _mm_popcnt_u32
+#  else
+#    include <popcntintrin.h>
+#    if defined(__x86_64__)
+#      define CV_POPCNT_U64 __builtin_popcountll
+#    endif
+#    define CV_POPCNT_U32 __builtin_popcount
+#  endif
+#  define CV_POPCNT 1
+#endif
+#ifdef CV_CPU_COMPILE_AVX
+#  include <immintrin.h>
+#  define CV_AVX 1
+#endif
+#ifdef CV_CPU_COMPILE_FP16
+#  if defined(__arm__) || defined(__aarch64__) || defined(_M_ARM) || defined(_M_ARM64)
+#    include <arm_neon.h>
+#  else
+#    include <immintrin.h>
+#  endif
+#  define CV_FP16 1
+#endif
+#ifdef CV_CPU_COMPILE_AVX2
+#  include <immintrin.h>
+#  define CV_AVX2 1
+#endif
+#ifdef CV_CPU_COMPILE_AVX_512F
+#  include <immintrin.h>
+#  define CV_AVX_512F 1
+#endif
+#ifdef CV_CPU_COMPILE_AVX512_COMMON
+#  define CV_AVX512_COMMON 1
+#  define CV_AVX_512CD 1
+#endif
+#ifdef CV_CPU_COMPILE_AVX512_KNL
+#  define CV_AVX512_KNL 1
+#  define CV_AVX_512ER 1
+#  define CV_AVX_512PF 1
+#endif
+#ifdef CV_CPU_COMPILE_AVX512_KNM
+#  define CV_AVX512_KNM 1
+#  define CV_AVX_5124FMAPS 1
+#  define CV_AVX_5124VNNIW 1
+#  define CV_AVX_512VPOPCNTDQ 1
+#endif
+#ifdef CV_CPU_COMPILE_AVX512_SKX
+#  define CV_AVX512_SKX 1
+#  define CV_AVX_512VL 1
+#  define CV_AVX_512BW 1
+#  define CV_AVX_512DQ 1
+#endif
+#ifdef CV_CPU_COMPILE_AVX512_CNL
+#  define CV_AVX512_CNL 1
+#  define CV_AVX_512IFMA 1
+#  define CV_AVX_512VBMI 1
+#endif
+#ifdef CV_CPU_COMPILE_AVX512_CLX
+#  define CV_AVX512_CLX 1
+#  define CV_AVX_512VNNI 1
+#endif
+#ifdef CV_CPU_COMPILE_AVX512_ICL
+#  define CV_AVX512_ICL 1
+#  undef CV_AVX_512IFMA
+#  define CV_AVX_512IFMA 1
+#  undef CV_AVX_512VBMI
+#  define CV_AVX_512VBMI 1
+#  undef CV_AVX_512VNNI
+#  define CV_AVX_512VNNI 1
+#  define CV_AVX_512VBMI2 1
+#  define CV_AVX_512BITALG 1
+#  define CV_AVX_512VPOPCNTDQ 1
+#endif
+#ifdef CV_CPU_COMPILE_FMA3
+#  define CV_FMA3 1
+#endif
+
+#if defined _WIN32 && (defined(_M_ARM) || defined(_M_ARM64)) && (defined(CV_CPU_COMPILE_NEON) || !defined(_MSC_VER))
+# include <Intrin.h>
+# include <arm_neon.h>
+# define CV_NEON 1
+#elif defined(__ARM_NEON__) || (defined (__ARM_NEON) && defined(__aarch64__))
+#  include <arm_neon.h>
+#  define CV_NEON 1
+#endif
+
+#if defined(__ARM_NEON__) || defined(__aarch64__)
+#  include <arm_neon.h>
+#endif
+
+#ifdef CV_CPU_COMPILE_VSX
+#  include <altivec.h>
+#  undef vector
+#  undef pixel
+#  undef bool
+#  define CV_VSX 1
+#endif
+
+#ifdef CV_CPU_COMPILE_VSX3
+#  define CV_VSX3 1
+#endif
+
+#ifdef CV_CPU_COMPILE_MSA
+#  include "hal/msa_macros.h"
+#  define CV_MSA 1
+#endif
+
+#ifdef __EMSCRIPTEN__
+#  define CV_WASM_SIMD 1
+#  include <wasm_simd128.h>
+#endif
+
+#endif // CV_ENABLE_INTRINSICS && !CV_DISABLE_OPTIMIZATION && !__CUDACC__
+
+#if defined CV_CPU_COMPILE_AVX && !defined CV_CPU_BASELINE_COMPILE_AVX
+struct VZeroUpperGuard {
+#ifdef __GNUC__
+    __attribute__((always_inline))
+#endif
+    inline VZeroUpperGuard() { _mm256_zeroupper(); }
+#ifdef __GNUC__
+    __attribute__((always_inline))
+#endif
+    inline ~VZeroUpperGuard() { _mm256_zeroupper(); }
+};
+#define __CV_AVX_GUARD VZeroUpperGuard __vzeroupper_guard; CV_UNUSED(__vzeroupper_guard);
+#endif
+
+#ifdef __CV_AVX_GUARD
+#define CV_AVX_GUARD __CV_AVX_GUARD
+#else
+#define CV_AVX_GUARD
+#endif
+
+#endif // __OPENCV_BUILD
+
+
+
+#if !defined __OPENCV_BUILD /* Compatibility code */ \
+    && !defined __CUDACC__ /* do not include SSE/AVX/NEON headers for NVCC compiler */
+#if defined __SSE2__ || defined _M_X64 || (defined _M_IX86_FP && _M_IX86_FP >= 2)
+#  include <emmintrin.h>
+#  define CV_MMX 1
+#  define CV_SSE 1
+#  define CV_SSE2 1
+#elif defined _WIN32 && (defined(_M_ARM) || defined(_M_ARM64)) && (defined(CV_CPU_COMPILE_NEON) || !defined(_MSC_VER))
+# include <Intrin.h>
+# include <arm_neon.h>
+# define CV_NEON 1
+#elif defined(__ARM_NEON__) || (defined (__ARM_NEON) && defined(__aarch64__))
+#  include <arm_neon.h>
+#  define CV_NEON 1
+#elif defined(__VSX__) && defined(__PPC64__) && defined(__LITTLE_ENDIAN__)
+#  include <altivec.h>
+#  undef vector
+#  undef pixel
+#  undef bool
+#  define CV_VSX 1
+#endif
+
+#endif // !__OPENCV_BUILD && !__CUDACC (Compatibility code)
+
+
+
+#ifndef CV_MMX
+#  define CV_MMX 0
+#endif
+#ifndef CV_SSE
+#  define CV_SSE 0
+#endif
+#ifndef CV_SSE2
+#  define CV_SSE2 0
+#endif
+#ifndef CV_SSE3
+#  define CV_SSE3 0
+#endif
+#ifndef CV_SSSE3
+#  define CV_SSSE3 0
+#endif
+#ifndef CV_SSE4_1
+#  define CV_SSE4_1 0
+#endif
+#ifndef CV_SSE4_2
+#  define CV_SSE4_2 0
+#endif
+#ifndef CV_POPCNT
+#  define CV_POPCNT 0
+#endif
+#ifndef CV_AVX
+#  define CV_AVX 0
+#endif
+#ifndef CV_FP16
+#  define CV_FP16 0
+#endif
+#ifndef CV_AVX2
+#  define CV_AVX2 0
+#endif
+#ifndef CV_FMA3
+#  define CV_FMA3 0
+#endif
+#ifndef CV_AVX_512F
+#  define CV_AVX_512F 0
+#endif
+#ifndef CV_AVX_512BW
+#  define CV_AVX_512BW 0
+#endif
+#ifndef CV_AVX_512CD
+#  define CV_AVX_512CD 0
+#endif
+#ifndef CV_AVX_512DQ
+#  define CV_AVX_512DQ 0
+#endif
+#ifndef CV_AVX_512ER
+#  define CV_AVX_512ER 0
+#endif
+#ifndef CV_AVX_512IFMA
+#  define CV_AVX_512IFMA 0
+#endif
+#define CV_AVX_512IFMA512 CV_AVX_512IFMA // deprecated
+#ifndef CV_AVX_512PF
+#  define CV_AVX_512PF 0
+#endif
+#ifndef CV_AVX_512VBMI
+#  define CV_AVX_512VBMI 0
+#endif
+#ifndef CV_AVX_512VL
+#  define CV_AVX_512VL 0
+#endif
+#ifndef CV_AVX_5124FMAPS
+#  define CV_AVX_5124FMAPS 0
+#endif
+#ifndef CV_AVX_5124VNNIW
+#  define CV_AVX_5124VNNIW 0
+#endif
+#ifndef CV_AVX_512VPOPCNTDQ
+#  define CV_AVX_512VPOPCNTDQ 0
+#endif
+#ifndef CV_AVX_512VNNI
+#  define CV_AVX_512VNNI 0
+#endif
+#ifndef CV_AVX_512VBMI2
+#  define CV_AVX_512VBMI2 0
+#endif
+#ifndef CV_AVX_512BITALG
+#  define CV_AVX_512BITALG 0
+#endif
+#ifndef CV_AVX512_COMMON
+#  define CV_AVX512_COMMON 0
+#endif
+#ifndef CV_AVX512_KNL
+#  define CV_AVX512_KNL 0
+#endif
+#ifndef CV_AVX512_KNM
+#  define CV_AVX512_KNM 0
+#endif
+#ifndef CV_AVX512_SKX
+#  define CV_AVX512_SKX 0
+#endif
+#ifndef CV_AVX512_CNL
+#  define CV_AVX512_CNL 0
+#endif
+#ifndef CV_AVX512_CLX
+#  define CV_AVX512_CLX 0
+#endif
+#ifndef CV_AVX512_ICL
+#  define CV_AVX512_ICL 0
+#endif
+
+#ifndef CV_NEON
+#  define CV_NEON 0
+#endif
+
+#ifndef CV_VSX
+#  define CV_VSX 0
+#endif
+
+#ifndef CV_VSX3
+#  define CV_VSX3 0
+#endif
+
+#ifndef CV_MSA
+#  define CV_MSA 0
+#endif
+
+#ifndef CV_WASM_SIMD
+#  define CV_WASM_SIMD 0
+#endif

model_zoo/official/lite/app/src/main/cpp/include/opencv2/core/cv_cpu_helper.h

@@ -0,0 +1,487 @@
+// AUTOGENERATED, DO NOT EDIT
+
+#if !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_COMPILE_SSE
+#  define CV_TRY_SSE 1
+#  define CV_CPU_FORCE_SSE 1
+#  define CV_CPU_HAS_SUPPORT_SSE 1
+#  define CV_CPU_CALL_SSE(fn, args) return (cpu_baseline::fn args)
+#  define CV_CPU_CALL_SSE_(fn, args) return (opt_SSE::fn args)
+#elif !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_DISPATCH_COMPILE_SSE
+#  define CV_TRY_SSE 1
+#  define CV_CPU_FORCE_SSE 0
+#  define CV_CPU_HAS_SUPPORT_SSE (cv::checkHardwareSupport(CV_CPU_SSE))
+#  define CV_CPU_CALL_SSE(fn, args) if (CV_CPU_HAS_SUPPORT_SSE) return (opt_SSE::fn args)
+#  define CV_CPU_CALL_SSE_(fn, args) if (CV_CPU_HAS_SUPPORT_SSE) return (opt_SSE::fn args)
+#else
+#  define CV_TRY_SSE 0
+#  define CV_CPU_FORCE_SSE 0
+#  define CV_CPU_HAS_SUPPORT_SSE 0
+#  define CV_CPU_CALL_SSE(fn, args)
+#  define CV_CPU_CALL_SSE_(fn, args)
+#endif
+#define __CV_CPU_DISPATCH_CHAIN_SSE(fn, args, mode, ...)  CV_CPU_CALL_SSE(fn, args); __CV_EXPAND(__CV_CPU_DISPATCH_CHAIN_ ## mode(fn, args, __VA_ARGS__))
+
+#if !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_COMPILE_SSE2
+#  define CV_TRY_SSE2 1
+#  define CV_CPU_FORCE_SSE2 1
+#  define CV_CPU_HAS_SUPPORT_SSE2 1
+#  define CV_CPU_CALL_SSE2(fn, args) return (cpu_baseline::fn args)
+#  define CV_CPU_CALL_SSE2_(fn, args) return (opt_SSE2::fn args)
+#elif !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_DISPATCH_COMPILE_SSE2
+#  define CV_TRY_SSE2 1
+#  define CV_CPU_FORCE_SSE2 0
+#  define CV_CPU_HAS_SUPPORT_SSE2 (cv::checkHardwareSupport(CV_CPU_SSE2))
+#  define CV_CPU_CALL_SSE2(fn, args) if (CV_CPU_HAS_SUPPORT_SSE2) return (opt_SSE2::fn args)
+#  define CV_CPU_CALL_SSE2_(fn, args) if (CV_CPU_HAS_SUPPORT_SSE2) return (opt_SSE2::fn args)
+#else
+#  define CV_TRY_SSE2 0
+#  define CV_CPU_FORCE_SSE2 0
+#  define CV_CPU_HAS_SUPPORT_SSE2 0
+#  define CV_CPU_CALL_SSE2(fn, args)
+#  define CV_CPU_CALL_SSE2_(fn, args)
+#endif
+#define __CV_CPU_DISPATCH_CHAIN_SSE2(fn, args, mode, ...)  CV_CPU_CALL_SSE2(fn, args); __CV_EXPAND(__CV_CPU_DISPATCH_CHAIN_ ## mode(fn, args, __VA_ARGS__))
+
+#if !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_COMPILE_SSE3
+#  define CV_TRY_SSE3 1
+#  define CV_CPU_FORCE_SSE3 1
+#  define CV_CPU_HAS_SUPPORT_SSE3 1
+#  define CV_CPU_CALL_SSE3(fn, args) return (cpu_baseline::fn args)
+#  define CV_CPU_CALL_SSE3_(fn, args) return (opt_SSE3::fn args)
+#elif !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_DISPATCH_COMPILE_SSE3
+#  define CV_TRY_SSE3 1
+#  define CV_CPU_FORCE_SSE3 0
+#  define CV_CPU_HAS_SUPPORT_SSE3 (cv::checkHardwareSupport(CV_CPU_SSE3))
+#  define CV_CPU_CALL_SSE3(fn, args) if (CV_CPU_HAS_SUPPORT_SSE3) return (opt_SSE3::fn args)
+#  define CV_CPU_CALL_SSE3_(fn, args) if (CV_CPU_HAS_SUPPORT_SSE3) return (opt_SSE3::fn args)
+#else
+#  define CV_TRY_SSE3 0
+#  define CV_CPU_FORCE_SSE3 0
+#  define CV_CPU_HAS_SUPPORT_SSE3 0
+#  define CV_CPU_CALL_SSE3(fn, args)
+#  define CV_CPU_CALL_SSE3_(fn, args)
+#endif
+#define __CV_CPU_DISPATCH_CHAIN_SSE3(fn, args, mode, ...)  CV_CPU_CALL_SSE3(fn, args); __CV_EXPAND(__CV_CPU_DISPATCH_CHAIN_ ## mode(fn, args, __VA_ARGS__))
+
+#if !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_COMPILE_SSSE3
+#  define CV_TRY_SSSE3 1
+#  define CV_CPU_FORCE_SSSE3 1
+#  define CV_CPU_HAS_SUPPORT_SSSE3 1
+#  define CV_CPU_CALL_SSSE3(fn, args) return (cpu_baseline::fn args)
+#  define CV_CPU_CALL_SSSE3_(fn, args) return (opt_SSSE3::fn args)
+#elif !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_DISPATCH_COMPILE_SSSE3
+#  define CV_TRY_SSSE3 1
+#  define CV_CPU_FORCE_SSSE3 0
+#  define CV_CPU_HAS_SUPPORT_SSSE3 (cv::checkHardwareSupport(CV_CPU_SSSE3))
+#  define CV_CPU_CALL_SSSE3(fn, args) if (CV_CPU_HAS_SUPPORT_SSSE3) return (opt_SSSE3::fn args)
+#  define CV_CPU_CALL_SSSE3_(fn, args) if (CV_CPU_HAS_SUPPORT_SSSE3) return (opt_SSSE3::fn args)
+#else
+#  define CV_TRY_SSSE3 0
+#  define CV_CPU_FORCE_SSSE3 0
+#  define CV_CPU_HAS_SUPPORT_SSSE3 0
+#  define CV_CPU_CALL_SSSE3(fn, args)
+#  define CV_CPU_CALL_SSSE3_(fn, args)
+#endif
+#define __CV_CPU_DISPATCH_CHAIN_SSSE3(fn, args, mode, ...)  CV_CPU_CALL_SSSE3(fn, args); __CV_EXPAND(__CV_CPU_DISPATCH_CHAIN_ ## mode(fn, args, __VA_ARGS__))
+
+#if !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_COMPILE_SSE4_1
+#  define CV_TRY_SSE4_1 1
+#  define CV_CPU_FORCE_SSE4_1 1
+#  define CV_CPU_HAS_SUPPORT_SSE4_1 1
+#  define CV_CPU_CALL_SSE4_1(fn, args) return (cpu_baseline::fn args)
+#  define CV_CPU_CALL_SSE4_1_(fn, args) return (opt_SSE4_1::fn args)
+#elif !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_DISPATCH_COMPILE_SSE4_1
+#  define CV_TRY_SSE4_1 1
+#  define CV_CPU_FORCE_SSE4_1 0
+#  define CV_CPU_HAS_SUPPORT_SSE4_1 (cv::checkHardwareSupport(CV_CPU_SSE4_1))
+#  define CV_CPU_CALL_SSE4_1(fn, args) if (CV_CPU_HAS_SUPPORT_SSE4_1) return (opt_SSE4_1::fn args)
+#  define CV_CPU_CALL_SSE4_1_(fn, args) if (CV_CPU_HAS_SUPPORT_SSE4_1) return (opt_SSE4_1::fn args)
+#else
+#  define CV_TRY_SSE4_1 0
+#  define CV_CPU_FORCE_SSE4_1 0
+#  define CV_CPU_HAS_SUPPORT_SSE4_1 0
+#  define CV_CPU_CALL_SSE4_1(fn, args)
+#  define CV_CPU_CALL_SSE4_1_(fn, args)
+#endif
+#define __CV_CPU_DISPATCH_CHAIN_SSE4_1(fn, args, mode, ...)  CV_CPU_CALL_SSE4_1(fn, args); __CV_EXPAND(__CV_CPU_DISPATCH_CHAIN_ ## mode(fn, args, __VA_ARGS__))
+
+#if !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_COMPILE_SSE4_2
+#  define CV_TRY_SSE4_2 1
+#  define CV_CPU_FORCE_SSE4_2 1
+#  define CV_CPU_HAS_SUPPORT_SSE4_2 1
+#  define CV_CPU_CALL_SSE4_2(fn, args) return (cpu_baseline::fn args)
+#  define CV_CPU_CALL_SSE4_2_(fn, args) return (opt_SSE4_2::fn args)
+#elif !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_DISPATCH_COMPILE_SSE4_2
+#  define CV_TRY_SSE4_2 1
+#  define CV_CPU_FORCE_SSE4_2 0
+#  define CV_CPU_HAS_SUPPORT_SSE4_2 (cv::checkHardwareSupport(CV_CPU_SSE4_2))
+#  define CV_CPU_CALL_SSE4_2(fn, args) if (CV_CPU_HAS_SUPPORT_SSE4_2) return (opt_SSE4_2::fn args)
+#  define CV_CPU_CALL_SSE4_2_(fn, args) if (CV_CPU_HAS_SUPPORT_SSE4_2) return (opt_SSE4_2::fn args)
+#else
+#  define CV_TRY_SSE4_2 0
+#  define CV_CPU_FORCE_SSE4_2 0
+#  define CV_CPU_HAS_SUPPORT_SSE4_2 0
+#  define CV_CPU_CALL_SSE4_2(fn, args)
+#  define CV_CPU_CALL_SSE4_2_(fn, args)
+#endif
+#define __CV_CPU_DISPATCH_CHAIN_SSE4_2(fn, args, mode, ...)  CV_CPU_CALL_SSE4_2(fn, args); __CV_EXPAND(__CV_CPU_DISPATCH_CHAIN_ ## mode(fn, args, __VA_ARGS__))
+
+#if !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_COMPILE_POPCNT
+#  define CV_TRY_POPCNT 1
+#  define CV_CPU_FORCE_POPCNT 1
+#  define CV_CPU_HAS_SUPPORT_POPCNT 1
+#  define CV_CPU_CALL_POPCNT(fn, args) return (cpu_baseline::fn args)
+#  define CV_CPU_CALL_POPCNT_(fn, args) return (opt_POPCNT::fn args)
+#elif !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_DISPATCH_COMPILE_POPCNT
+#  define CV_TRY_POPCNT 1
+#  define CV_CPU_FORCE_POPCNT 0
+#  define CV_CPU_HAS_SUPPORT_POPCNT (cv::checkHardwareSupport(CV_CPU_POPCNT))
+#  define CV_CPU_CALL_POPCNT(fn, args) if (CV_CPU_HAS_SUPPORT_POPCNT) return (opt_POPCNT::fn args)
+#  define CV_CPU_CALL_POPCNT_(fn, args) if (CV_CPU_HAS_SUPPORT_POPCNT) return (opt_POPCNT::fn args)
+#else
+#  define CV_TRY_POPCNT 0
+#  define CV_CPU_FORCE_POPCNT 0
+#  define CV_CPU_HAS_SUPPORT_POPCNT 0
+#  define CV_CPU_CALL_POPCNT(fn, args)
+#  define CV_CPU_CALL_POPCNT_(fn, args)
+#endif
+#define __CV_CPU_DISPATCH_CHAIN_POPCNT(fn, args, mode, ...)  CV_CPU_CALL_POPCNT(fn, args); __CV_EXPAND(__CV_CPU_DISPATCH_CHAIN_ ## mode(fn, args, __VA_ARGS__))
+
+#if !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_COMPILE_AVX
+#  define CV_TRY_AVX 1
+#  define CV_CPU_FORCE_AVX 1
+#  define CV_CPU_HAS_SUPPORT_AVX 1
+#  define CV_CPU_CALL_AVX(fn, args) return (cpu_baseline::fn args)
+#  define CV_CPU_CALL_AVX_(fn, args) return (opt_AVX::fn args)
+#elif !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_DISPATCH_COMPILE_AVX
+#  define CV_TRY_AVX 1
+#  define CV_CPU_FORCE_AVX 0
+#  define CV_CPU_HAS_SUPPORT_AVX (cv::checkHardwareSupport(CV_CPU_AVX))
+#  define CV_CPU_CALL_AVX(fn, args) if (CV_CPU_HAS_SUPPORT_AVX) return (opt_AVX::fn args)
+#  define CV_CPU_CALL_AVX_(fn, args) if (CV_CPU_HAS_SUPPORT_AVX) return (opt_AVX::fn args)
+#else
+#  define CV_TRY_AVX 0
+#  define CV_CPU_FORCE_AVX 0
+#  define CV_CPU_HAS_SUPPORT_AVX 0
+#  define CV_CPU_CALL_AVX(fn, args)
+#  define CV_CPU_CALL_AVX_(fn, args)
+#endif
+#define __CV_CPU_DISPATCH_CHAIN_AVX(fn, args, mode, ...)  CV_CPU_CALL_AVX(fn, args); __CV_EXPAND(__CV_CPU_DISPATCH_CHAIN_ ## mode(fn, args, __VA_ARGS__))
+
+#if !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_COMPILE_FP16
+#  define CV_TRY_FP16 1
+#  define CV_CPU_FORCE_FP16 1
+#  define CV_CPU_HAS_SUPPORT_FP16 1
+#  define CV_CPU_CALL_FP16(fn, args) return (cpu_baseline::fn args)
+#  define CV_CPU_CALL_FP16_(fn, args) return (opt_FP16::fn args)
+#elif !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_DISPATCH_COMPILE_FP16
+#  define CV_TRY_FP16 1
+#  define CV_CPU_FORCE_FP16 0
+#  define CV_CPU_HAS_SUPPORT_FP16 (cv::checkHardwareSupport(CV_CPU_FP16))
+#  define CV_CPU_CALL_FP16(fn, args) if (CV_CPU_HAS_SUPPORT_FP16) return (opt_FP16::fn args)
+#  define CV_CPU_CALL_FP16_(fn, args) if (CV_CPU_HAS_SUPPORT_FP16) return (opt_FP16::fn args)
+#else
+#  define CV_TRY_FP16 0
+#  define CV_CPU_FORCE_FP16 0
+#  define CV_CPU_HAS_SUPPORT_FP16 0
+#  define CV_CPU_CALL_FP16(fn, args)
+#  define CV_CPU_CALL_FP16_(fn, args)
+#endif
+#define __CV_CPU_DISPATCH_CHAIN_FP16(fn, args, mode, ...)  CV_CPU_CALL_FP16(fn, args); __CV_EXPAND(__CV_CPU_DISPATCH_CHAIN_ ## mode(fn, args, __VA_ARGS__))
+
+#if !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_COMPILE_AVX2
+#  define CV_TRY_AVX2 1
+#  define CV_CPU_FORCE_AVX2 1
+#  define CV_CPU_HAS_SUPPORT_AVX2 1
+#  define CV_CPU_CALL_AVX2(fn, args) return (cpu_baseline::fn args)
+#  define CV_CPU_CALL_AVX2_(fn, args) return (opt_AVX2::fn args)
+#elif !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_DISPATCH_COMPILE_AVX2
+#  define CV_TRY_AVX2 1
+#  define CV_CPU_FORCE_AVX2 0
+#  define CV_CPU_HAS_SUPPORT_AVX2 (cv::checkHardwareSupport(CV_CPU_AVX2))
+#  define CV_CPU_CALL_AVX2(fn, args) if (CV_CPU_HAS_SUPPORT_AVX2) return (opt_AVX2::fn args)
+#  define CV_CPU_CALL_AVX2_(fn, args) if (CV_CPU_HAS_SUPPORT_AVX2) return (opt_AVX2::fn args)
+#else
+#  define CV_TRY_AVX2 0
+#  define CV_CPU_FORCE_AVX2 0
+#  define CV_CPU_HAS_SUPPORT_AVX2 0
+#  define CV_CPU_CALL_AVX2(fn, args)
+#  define CV_CPU_CALL_AVX2_(fn, args)
+#endif
+#define __CV_CPU_DISPATCH_CHAIN_AVX2(fn, args, mode, ...)  CV_CPU_CALL_AVX2(fn, args); __CV_EXPAND(__CV_CPU_DISPATCH_CHAIN_ ## mode(fn, args, __VA_ARGS__))
+
+#if !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_COMPILE_FMA3
+#  define CV_TRY_FMA3 1
+#  define CV_CPU_FORCE_FMA3 1
+#  define CV_CPU_HAS_SUPPORT_FMA3 1
+#  define CV_CPU_CALL_FMA3(fn, args) return (cpu_baseline::fn args)
+#  define CV_CPU_CALL_FMA3_(fn, args) return (opt_FMA3::fn args)
+#elif !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_DISPATCH_COMPILE_FMA3
+#  define CV_TRY_FMA3 1
+#  define CV_CPU_FORCE_FMA3 0
+#  define CV_CPU_HAS_SUPPORT_FMA3 (cv::checkHardwareSupport(CV_CPU_FMA3))
+#  define CV_CPU_CALL_FMA3(fn, args) if (CV_CPU_HAS_SUPPORT_FMA3) return (opt_FMA3::fn args)
+#  define CV_CPU_CALL_FMA3_(fn, args) if (CV_CPU_HAS_SUPPORT_FMA3) return (opt_FMA3::fn args)
+#else
+#  define CV_TRY_FMA3 0
+#  define CV_CPU_FORCE_FMA3 0
+#  define CV_CPU_HAS_SUPPORT_FMA3 0
+#  define CV_CPU_CALL_FMA3(fn, args)
+#  define CV_CPU_CALL_FMA3_(fn, args)
+#endif
+#define __CV_CPU_DISPATCH_CHAIN_FMA3(fn, args, mode, ...)  CV_CPU_CALL_FMA3(fn, args); __CV_EXPAND(__CV_CPU_DISPATCH_CHAIN_ ## mode(fn, args, __VA_ARGS__))
+
+#if !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_COMPILE_AVX_512F
+#  define CV_TRY_AVX_512F 1
+#  define CV_CPU_FORCE_AVX_512F 1
+#  define CV_CPU_HAS_SUPPORT_AVX_512F 1
+#  define CV_CPU_CALL_AVX_512F(fn, args) return (cpu_baseline::fn args)
+#  define CV_CPU_CALL_AVX_512F_(fn, args) return (opt_AVX_512F::fn args)
+#elif !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_DISPATCH_COMPILE_AVX_512F
+#  define CV_TRY_AVX_512F 1
+#  define CV_CPU_FORCE_AVX_512F 0
+#  define CV_CPU_HAS_SUPPORT_AVX_512F (cv::checkHardwareSupport(CV_CPU_AVX_512F))
+#  define CV_CPU_CALL_AVX_512F(fn, args) if (CV_CPU_HAS_SUPPORT_AVX_512F) return (opt_AVX_512F::fn args)
+#  define CV_CPU_CALL_AVX_512F_(fn, args) if (CV_CPU_HAS_SUPPORT_AVX_512F) return (opt_AVX_512F::fn args)
+#else
+#  define CV_TRY_AVX_512F 0
+#  define CV_CPU_FORCE_AVX_512F 0
+#  define CV_CPU_HAS_SUPPORT_AVX_512F 0
+#  define CV_CPU_CALL_AVX_512F(fn, args)
+#  define CV_CPU_CALL_AVX_512F_(fn, args)
+#endif
+#define __CV_CPU_DISPATCH_CHAIN_AVX_512F(fn, args, mode, ...)  CV_CPU_CALL_AVX_512F(fn, args); __CV_EXPAND(__CV_CPU_DISPATCH_CHAIN_ ## mode(fn, args, __VA_ARGS__))
+
+#if !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_COMPILE_AVX512_COMMON
+#  define CV_TRY_AVX512_COMMON 1
+#  define CV_CPU_FORCE_AVX512_COMMON 1
+#  define CV_CPU_HAS_SUPPORT_AVX512_COMMON 1
+#  define CV_CPU_CALL_AVX512_COMMON(fn, args) return (cpu_baseline::fn args)
+#  define CV_CPU_CALL_AVX512_COMMON_(fn, args) return (opt_AVX512_COMMON::fn args)
+#elif !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_DISPATCH_COMPILE_AVX512_COMMON
+#  define CV_TRY_AVX512_COMMON 1
+#  define CV_CPU_FORCE_AVX512_COMMON 0
+#  define CV_CPU_HAS_SUPPORT_AVX512_COMMON (cv::checkHardwareSupport(CV_CPU_AVX512_COMMON))
+#  define CV_CPU_CALL_AVX512_COMMON(fn, args) if (CV_CPU_HAS_SUPPORT_AVX512_COMMON) return (opt_AVX512_COMMON::fn args)
+#  define CV_CPU_CALL_AVX512_COMMON_(fn, args) if (CV_CPU_HAS_SUPPORT_AVX512_COMMON) return (opt_AVX512_COMMON::fn args)
+#else
+#  define CV_TRY_AVX512_COMMON 0
+#  define CV_CPU_FORCE_AVX512_COMMON 0
+#  define CV_CPU_HAS_SUPPORT_AVX512_COMMON 0
+#  define CV_CPU_CALL_AVX512_COMMON(fn, args)
+#  define CV_CPU_CALL_AVX512_COMMON_(fn, args)
+#endif
+#define __CV_CPU_DISPATCH_CHAIN_AVX512_COMMON(fn, args, mode, ...)  CV_CPU_CALL_AVX512_COMMON(fn, args); __CV_EXPAND(__CV_CPU_DISPATCH_CHAIN_ ## mode(fn, args, __VA_ARGS__))
+
+#if !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_COMPILE_AVX512_KNL
+#  define CV_TRY_AVX512_KNL 1
+#  define CV_CPU_FORCE_AVX512_KNL 1
+#  define CV_CPU_HAS_SUPPORT_AVX512_KNL 1
+#  define CV_CPU_CALL_AVX512_KNL(fn, args) return (cpu_baseline::fn args)
+#  define CV_CPU_CALL_AVX512_KNL_(fn, args) return (opt_AVX512_KNL::fn args)
+#elif !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_DISPATCH_COMPILE_AVX512_KNL
+#  define CV_TRY_AVX512_KNL 1
+#  define CV_CPU_FORCE_AVX512_KNL 0
+#  define CV_CPU_HAS_SUPPORT_AVX512_KNL (cv::checkHardwareSupport(CV_CPU_AVX512_KNL))
+#  define CV_CPU_CALL_AVX512_KNL(fn, args) if (CV_CPU_HAS_SUPPORT_AVX512_KNL) return (opt_AVX512_KNL::fn args)
+#  define CV_CPU_CALL_AVX512_KNL_(fn, args) if (CV_CPU_HAS_SUPPORT_AVX512_KNL) return (opt_AVX512_KNL::fn args)
+#else
+#  define CV_TRY_AVX512_KNL 0
+#  define CV_CPU_FORCE_AVX512_KNL 0
+#  define CV_CPU_HAS_SUPPORT_AVX512_KNL 0
+#  define CV_CPU_CALL_AVX512_KNL(fn, args)
+#  define CV_CPU_CALL_AVX512_KNL_(fn, args)
+#endif
+#define __CV_CPU_DISPATCH_CHAIN_AVX512_KNL(fn, args, mode, ...)  CV_CPU_CALL_AVX512_KNL(fn, args); __CV_EXPAND(__CV_CPU_DISPATCH_CHAIN_ ## mode(fn, args, __VA_ARGS__))
+
+#if !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_COMPILE_AVX512_KNM
+#  define CV_TRY_AVX512_KNM 1
+#  define CV_CPU_FORCE_AVX512_KNM 1
+#  define CV_CPU_HAS_SUPPORT_AVX512_KNM 1
+#  define CV_CPU_CALL_AVX512_KNM(fn, args) return (cpu_baseline::fn args)
+#  define CV_CPU_CALL_AVX512_KNM_(fn, args) return (opt_AVX512_KNM::fn args)
+#elif !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_DISPATCH_COMPILE_AVX512_KNM
+#  define CV_TRY_AVX512_KNM 1
+#  define CV_CPU_FORCE_AVX512_KNM 0
+#  define CV_CPU_HAS_SUPPORT_AVX512_KNM (cv::checkHardwareSupport(CV_CPU_AVX512_KNM))
+#  define CV_CPU_CALL_AVX512_KNM(fn, args) if (CV_CPU_HAS_SUPPORT_AVX512_KNM) return (opt_AVX512_KNM::fn args)
+#  define CV_CPU_CALL_AVX512_KNM_(fn, args) if (CV_CPU_HAS_SUPPORT_AVX512_KNM) return (opt_AVX512_KNM::fn args)
+#else
+#  define CV_TRY_AVX512_KNM 0
+#  define CV_CPU_FORCE_AVX512_KNM 0
+#  define CV_CPU_HAS_SUPPORT_AVX512_KNM 0
+#  define CV_CPU_CALL_AVX512_KNM(fn, args)
+#  define CV_CPU_CALL_AVX512_KNM_(fn, args)
+#endif
+#define __CV_CPU_DISPATCH_CHAIN_AVX512_KNM(fn, args, mode, ...)  CV_CPU_CALL_AVX512_KNM(fn, args); __CV_EXPAND(__CV_CPU_DISPATCH_CHAIN_ ## mode(fn, args, __VA_ARGS__))
+
+#if !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_COMPILE_AVX512_SKX
+#  define CV_TRY_AVX512_SKX 1
+#  define CV_CPU_FORCE_AVX512_SKX 1
+#  define CV_CPU_HAS_SUPPORT_AVX512_SKX 1
+#  define CV_CPU_CALL_AVX512_SKX(fn, args) return (cpu_baseline::fn args)
+#  define CV_CPU_CALL_AVX512_SKX_(fn, args) return (opt_AVX512_SKX::fn args)
+#elif !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_DISPATCH_COMPILE_AVX512_SKX
+#  define CV_TRY_AVX512_SKX 1
+#  define CV_CPU_FORCE_AVX512_SKX 0
+#  define CV_CPU_HAS_SUPPORT_AVX512_SKX (cv::checkHardwareSupport(CV_CPU_AVX512_SKX))
+#  define CV_CPU_CALL_AVX512_SKX(fn, args) if (CV_CPU_HAS_SUPPORT_AVX512_SKX) return (opt_AVX512_SKX::fn args)
+#  define CV_CPU_CALL_AVX512_SKX_(fn, args) if (CV_CPU_HAS_SUPPORT_AVX512_SKX) return (opt_AVX512_SKX::fn args)
+#else
+#  define CV_TRY_AVX512_SKX 0
+#  define CV_CPU_FORCE_AVX512_SKX 0
+#  define CV_CPU_HAS_SUPPORT_AVX512_SKX 0
+#  define CV_CPU_CALL_AVX512_SKX(fn, args)
+#  define CV_CPU_CALL_AVX512_SKX_(fn, args)
+#endif
+#define __CV_CPU_DISPATCH_CHAIN_AVX512_SKX(fn, args, mode, ...)  CV_CPU_CALL_AVX512_SKX(fn, args); __CV_EXPAND(__CV_CPU_DISPATCH_CHAIN_ ## mode(fn, args, __VA_ARGS__))
+
+#if !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_COMPILE_AVX512_CNL
+#  define CV_TRY_AVX512_CNL 1
+#  define CV_CPU_FORCE_AVX512_CNL 1
+#  define CV_CPU_HAS_SUPPORT_AVX512_CNL 1
+#  define CV_CPU_CALL_AVX512_CNL(fn, args) return (cpu_baseline::fn args)
+#  define CV_CPU_CALL_AVX512_CNL_(fn, args) return (opt_AVX512_CNL::fn args)
+#elif !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_DISPATCH_COMPILE_AVX512_CNL
+#  define CV_TRY_AVX512_CNL 1
+#  define CV_CPU_FORCE_AVX512_CNL 0
+#  define CV_CPU_HAS_SUPPORT_AVX512_CNL (cv::checkHardwareSupport(CV_CPU_AVX512_CNL))
+#  define CV_CPU_CALL_AVX512_CNL(fn, args) if (CV_CPU_HAS_SUPPORT_AVX512_CNL) return (opt_AVX512_CNL::fn args)
+#  define CV_CPU_CALL_AVX512_CNL_(fn, args) if (CV_CPU_HAS_SUPPORT_AVX512_CNL) return (opt_AVX512_CNL::fn args)
+#else
+#  define CV_TRY_AVX512_CNL 0
+#  define CV_CPU_FORCE_AVX512_CNL 0
+#  define CV_CPU_HAS_SUPPORT_AVX512_CNL 0
+#  define CV_CPU_CALL_AVX512_CNL(fn, args)
+#  define CV_CPU_CALL_AVX512_CNL_(fn, args)
+#endif
+#define __CV_CPU_DISPATCH_CHAIN_AVX512_CNL(fn, args, mode, ...)  CV_CPU_CALL_AVX512_CNL(fn, args); __CV_EXPAND(__CV_CPU_DISPATCH_CHAIN_ ## mode(fn, args, __VA_ARGS__))
+
+#if !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_COMPILE_AVX512_CLX
+#  define CV_TRY_AVX512_CLX 1
+#  define CV_CPU_FORCE_AVX512_CLX 1
+#  define CV_CPU_HAS_SUPPORT_AVX512_CLX 1
+#  define CV_CPU_CALL_AVX512_CLX(fn, args) return (cpu_baseline::fn args)
+#  define CV_CPU_CALL_AVX512_CLX_(fn, args) return (opt_AVX512_CLX::fn args)
+#elif !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_DISPATCH_COMPILE_AVX512_CLX
+#  define CV_TRY_AVX512_CLX 1
+#  define CV_CPU_FORCE_AVX512_CLX 0
+#  define CV_CPU_HAS_SUPPORT_AVX512_CLX (cv::checkHardwareSupport(CV_CPU_AVX512_CLX))
+#  define CV_CPU_CALL_AVX512_CLX(fn, args) if (CV_CPU_HAS_SUPPORT_AVX512_CLX) return (opt_AVX512_CLX::fn args)
+#  define CV_CPU_CALL_AVX512_CLX_(fn, args) if (CV_CPU_HAS_SUPPORT_AVX512_CLX) return (opt_AVX512_CLX::fn args)
+#else
+#  define CV_TRY_AVX512_CLX 0
+#  define CV_CPU_FORCE_AVX512_CLX 0
+#  define CV_CPU_HAS_SUPPORT_AVX512_CLX 0
+#  define CV_CPU_CALL_AVX512_CLX(fn, args)
+#  define CV_CPU_CALL_AVX512_CLX_(fn, args)
+#endif
+#define __CV_CPU_DISPATCH_CHAIN_AVX512_CLX(fn, args, mode, ...)  CV_CPU_CALL_AVX512_CLX(fn, args); __CV_EXPAND(__CV_CPU_DISPATCH_CHAIN_ ## mode(fn, args, __VA_ARGS__))
+
+#if !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_COMPILE_AVX512_ICL
+#  define CV_TRY_AVX512_ICL 1
+#  define CV_CPU_FORCE_AVX512_ICL 1
+#  define CV_CPU_HAS_SUPPORT_AVX512_ICL 1
+#  define CV_CPU_CALL_AVX512_ICL(fn, args) return (cpu_baseline::fn args)
+#  define CV_CPU_CALL_AVX512_ICL_(fn, args) return (opt_AVX512_ICL::fn args)
+#elif !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_DISPATCH_COMPILE_AVX512_ICL
+#  define CV_TRY_AVX512_ICL 1
+#  define CV_CPU_FORCE_AVX512_ICL 0
+#  define CV_CPU_HAS_SUPPORT_AVX512_ICL (cv::checkHardwareSupport(CV_CPU_AVX512_ICL))
+#  define CV_CPU_CALL_AVX512_ICL(fn, args) if (CV_CPU_HAS_SUPPORT_AVX512_ICL) return (opt_AVX512_ICL::fn args)
+#  define CV_CPU_CALL_AVX512_ICL_(fn, args) if (CV_CPU_HAS_SUPPORT_AVX512_ICL) return (opt_AVX512_ICL::fn args)
+#else
+#  define CV_TRY_AVX512_ICL 0
+#  define CV_CPU_FORCE_AVX512_ICL 0
+#  define CV_CPU_HAS_SUPPORT_AVX512_ICL 0
+#  define CV_CPU_CALL_AVX512_ICL(fn, args)
+#  define CV_CPU_CALL_AVX512_ICL_(fn, args)
+#endif
+#define __CV_CPU_DISPATCH_CHAIN_AVX512_ICL(fn, args, mode, ...)  CV_CPU_CALL_AVX512_ICL(fn, args); __CV_EXPAND(__CV_CPU_DISPATCH_CHAIN_ ## mode(fn, args, __VA_ARGS__))
+
+#if !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_COMPILE_NEON
+#  define CV_TRY_NEON 1
+#  define CV_CPU_FORCE_NEON 1
+#  define CV_CPU_HAS_SUPPORT_NEON 1
+#  define CV_CPU_CALL_NEON(fn, args) return (cpu_baseline::fn args)
+#  define CV_CPU_CALL_NEON_(fn, args) return (opt_NEON::fn args)
+#elif !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_DISPATCH_COMPILE_NEON
+#  define CV_TRY_NEON 1
+#  define CV_CPU_FORCE_NEON 0
+#  define CV_CPU_HAS_SUPPORT_NEON (cv::checkHardwareSupport(CV_CPU_NEON))
+#  define CV_CPU_CALL_NEON(fn, args) if (CV_CPU_HAS_SUPPORT_NEON) return (opt_NEON::fn args)
+#  define CV_CPU_CALL_NEON_(fn, args) if (CV_CPU_HAS_SUPPORT_NEON) return (opt_NEON::fn args)
+#else
+#  define CV_TRY_NEON 0
+#  define CV_CPU_FORCE_NEON 0
+#  define CV_CPU_HAS_SUPPORT_NEON 0
+#  define CV_CPU_CALL_NEON(fn, args)
+#  define CV_CPU_CALL_NEON_(fn, args)
+#endif
+#define __CV_CPU_DISPATCH_CHAIN_NEON(fn, args, mode, ...)  CV_CPU_CALL_NEON(fn, args); __CV_EXPAND(__CV_CPU_DISPATCH_CHAIN_ ## mode(fn, args, __VA_ARGS__))
+
+#if !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_COMPILE_MSA
+#  define CV_TRY_MSA 1
+#  define CV_CPU_FORCE_MSA 1
+#  define CV_CPU_HAS_SUPPORT_MSA 1
+#  define CV_CPU_CALL_MSA(fn, args) return (cpu_baseline::fn args)
+#  define CV_CPU_CALL_MSA_(fn, args) return (opt_MSA::fn args)
+#elif !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_DISPATCH_COMPILE_MSA
+#  define CV_TRY_MSA 1
+#  define CV_CPU_FORCE_MSA 0
+#  define CV_CPU_HAS_SUPPORT_MSA (cv::checkHardwareSupport(CV_CPU_MSA))
+#  define CV_CPU_CALL_MSA(fn, args) if (CV_CPU_HAS_SUPPORT_MSA) return (opt_MSA::fn args)
+#  define CV_CPU_CALL_MSA_(fn, args) if (CV_CPU_HAS_SUPPORT_MSA) return (opt_MSA::fn args)
+#else
+#  define CV_TRY_MSA 0
+#  define CV_CPU_FORCE_MSA 0
+#  define CV_CPU_HAS_SUPPORT_MSA 0
+#  define CV_CPU_CALL_MSA(fn, args)
+#  define CV_CPU_CALL_MSA_(fn, args)
+#endif
+#define __CV_CPU_DISPATCH_CHAIN_MSA(fn, args, mode, ...)  CV_CPU_CALL_MSA(fn, args); __CV_EXPAND(__CV_CPU_DISPATCH_CHAIN_ ## mode(fn, args, __VA_ARGS__))
+
+#if !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_COMPILE_VSX
+#  define CV_TRY_VSX 1
+#  define CV_CPU_FORCE_VSX 1
+#  define CV_CPU_HAS_SUPPORT_VSX 1
+#  define CV_CPU_CALL_VSX(fn, args) return (cpu_baseline::fn args)
+#  define CV_CPU_CALL_VSX_(fn, args) return (opt_VSX::fn args)
+#elif !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_DISPATCH_COMPILE_VSX
+#  define CV_TRY_VSX 1
+#  define CV_CPU_FORCE_VSX 0
+#  define CV_CPU_HAS_SUPPORT_VSX (cv::checkHardwareSupport(CV_CPU_VSX))
+#  define CV_CPU_CALL_VSX(fn, args) if (CV_CPU_HAS_SUPPORT_VSX) return (opt_VSX::fn args)
+#  define CV_CPU_CALL_VSX_(fn, args) if (CV_CPU_HAS_SUPPORT_VSX) return (opt_VSX::fn args)
+#else
+#  define CV_TRY_VSX 0
+#  define CV_CPU_FORCE_VSX 0
+#  define CV_CPU_HAS_SUPPORT_VSX 0
+#  define CV_CPU_CALL_VSX(fn, args)
+#  define CV_CPU_CALL_VSX_(fn, args)
+#endif
+#define __CV_CPU_DISPATCH_CHAIN_VSX(fn, args, mode, ...)  CV_CPU_CALL_VSX(fn, args); __CV_EXPAND(__CV_CPU_DISPATCH_CHAIN_ ## mode(fn, args, __VA_ARGS__))
+
+#if !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_COMPILE_VSX3
+#  define CV_TRY_VSX3 1
+#  define CV_CPU_FORCE_VSX3 1
+#  define CV_CPU_HAS_SUPPORT_VSX3 1
+#  define CV_CPU_CALL_VSX3(fn, args) return (cpu_baseline::fn args)
+#  define CV_CPU_CALL_VSX3_(fn, args) return (opt_VSX3::fn args)
+#elif !defined CV_DISABLE_OPTIMIZATION && defined CV_ENABLE_INTRINSICS && defined CV_CPU_DISPATCH_COMPILE_VSX3
+#  define CV_TRY_VSX3 1
+#  define CV_CPU_FORCE_VSX3 0
+#  define CV_CPU_HAS_SUPPORT_VSX3 (cv::checkHardwareSupport(CV_CPU_VSX3))
+#  define CV_CPU_CALL_VSX3(fn, args) if (CV_CPU_HAS_SUPPORT_VSX3) return (opt_VSX3::fn args)
+#  define CV_CPU_CALL_VSX3_(fn, args) if (CV_CPU_HAS_SUPPORT_VSX3) return (opt_VSX3::fn args)
+#else
+#  define CV_TRY_VSX3 0
+#  define CV_CPU_FORCE_VSX3 0
+#  define CV_CPU_HAS_SUPPORT_VSX3 0
+#  define CV_CPU_CALL_VSX3(fn, args)
+#  define CV_CPU_CALL_VSX3_(fn, args)
+#endif
+#define __CV_CPU_DISPATCH_CHAIN_VSX3(fn, args, mode, ...)  CV_CPU_CALL_VSX3(fn, args); __CV_EXPAND(__CV_CPU_DISPATCH_CHAIN_ ## mode(fn, args, __VA_ARGS__))
+
+#define CV_CPU_CALL_BASELINE(fn, args) return (cpu_baseline::fn args)
+#define __CV_CPU_DISPATCH_CHAIN_BASELINE(fn, args, mode, ...)  CV_CPU_CALL_BASELINE(fn, args) /* last in sequence */

model_zoo/official/lite/app/src/main/cpp/include/opencv2/core/cvdef.h

@@ -0,0 +1,920 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Copyright (C) 2015, Itseez Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CORE_CVDEF_H
+#define OPENCV_CORE_CVDEF_H
+
+//! @addtogroup core_utils
+//! @{
+
+#ifdef OPENCV_INCLUDE_PORT_FILE  // User-provided header file with custom platform configuration
+#include OPENCV_INCLUDE_PORT_FILE
+#endif
+
+#if !defined CV_DOXYGEN && !defined CV_IGNORE_DEBUG_BUILD_GUARD
+#if (defined(_MSC_VER) && (defined(DEBUG) || defined(_DEBUG))) || \
+    (defined(_GLIBCXX_DEBUG) || defined(_GLIBCXX_DEBUG_PEDANTIC))
+// Guard to prevent using of binary incompatible binaries / runtimes
+// https://github.com/opencv/opencv/pull/9161
+#define CV__DEBUG_NS_BEGIN namespace debug_build_guard {
+#define CV__DEBUG_NS_END }
+namespace cv { namespace debug_build_guard { } using namespace debug_build_guard; }
+#endif
+#endif
+
+#ifndef CV__DEBUG_NS_BEGIN
+#define CV__DEBUG_NS_BEGIN
+#define CV__DEBUG_NS_END
+#endif
+
+
+#ifdef __OPENCV_BUILD
+#include "cvconfig.h"
+#endif
+
+#ifndef __CV_EXPAND
+#define __CV_EXPAND(x) x
+#endif
+
+#ifndef __CV_CAT
+#define __CV_CAT__(x, y) x ## y
+#define __CV_CAT_(x, y) __CV_CAT__(x, y)
+#define __CV_CAT(x, y) __CV_CAT_(x, y)
+#endif
+
+#define __CV_VA_NUM_ARGS_HELPER(_1, _2, _3, _4, _5, _6, _7, _8, _9, _10, N, ...) N
+#define __CV_VA_NUM_ARGS(...) __CV_EXPAND(__CV_VA_NUM_ARGS_HELPER(__VA_ARGS__, 10, 9, 8, 7, 6, 5, 4, 3, 2, 1, 0))
+
+#ifdef CV_Func
+// keep current value (through OpenCV port file)
+#elif defined __GNUC__ || (defined (__cpluscplus) && (__cpluscplus >= 201103))
+#define CV_Func __func__
+#elif defined __clang__ && (__clang_minor__ * 100 + __clang_major >= 305)
+#define CV_Func __func__
+#elif defined(__STDC_VERSION__) && (__STDC_VERSION >= 199901)
+#define CV_Func __func__
+#elif defined _MSC_VER
+#define CV_Func __FUNCTION__
+#elif defined(__INTEL_COMPILER) && (_INTEL_COMPILER >= 600)
+#define CV_Func __FUNCTION__
+#elif defined __IBMCPP__ && __IBMCPP__ >=500
+#define CV_Func __FUNCTION__
+#elif defined __BORLAND__ && (__BORLANDC__ >= 0x550)
+#define CV_Func __FUNC__
+#else
+#define CV_Func "<unknown>"
+#endif
+
+//! @cond IGNORED
+
+//////////////// static assert /////////////////
+#define CVAUX_CONCAT_EXP(a, b) a##b
+#define CVAUX_CONCAT(a, b) CVAUX_CONCAT_EXP(a,b)
+
+#if defined(__clang__)
+#  ifndef __has_extension
+#    define __has_extension __has_feature /* compatibility, for older versions of clang */
+#  endif
+#  if __has_extension(cxx_static_assert)
+#    define CV_StaticAssert(condition, reason)    static_assert((condition), reason " " #condition)
+#  elif __has_extension(c_static_assert)
+#    define CV_StaticAssert(condition, reason)    _Static_assert((condition), reason " " #condition)
+#  endif
+#elif defined(__GNUC__)
+#  if (defined(__GXX_EXPERIMENTAL_CXX0X__) || __cplusplus >= 201103L)
+#    define CV_StaticAssert(condition, reason)    static_assert((condition), reason " " #condition)
+#  endif
+#elif defined(_MSC_VER)
+#  if _MSC_VER >= 1600 /* MSVC 10 */
+#    define CV_StaticAssert(condition, reason)    static_assert((condition), reason " " #condition)
+#  endif
+#endif
+#ifndef CV_StaticAssert
+#  if !defined(__clang__) && defined(__GNUC__) && (__GNUC__*100 + __GNUC_MINOR__ > 302)
+#    define CV_StaticAssert(condition, reason) ({ extern int __attribute__((error("CV_StaticAssert: " reason " " #condition))) CV_StaticAssert(); ((condition) ? 0 : CV_StaticAssert()); })
+#  else
+namespace cv {
+     template <bool x> struct CV_StaticAssert_failed;
+     template <> struct CV_StaticAssert_failed<true> { enum { val = 1 }; };
+     template<int x> struct CV_StaticAssert_test {};
+}
+#    define CV_StaticAssert(condition, reason)\
+       typedef cv::CV_StaticAssert_test< sizeof(cv::CV_StaticAssert_failed< static_cast<bool>(condition) >) > CVAUX_CONCAT(CV_StaticAssert_failed_at_, __LINE__)
+#  endif
+#endif
+
+// Suppress warning "-Wdeprecated-declarations" / C4996
+#if defined(_MSC_VER)
+    #define CV_DO_PRAGMA(x) __pragma(x)
+#elif defined(__GNUC__)
+    #define CV_DO_PRAGMA(x) _Pragma (#x)
+#else
+    #define CV_DO_PRAGMA(x)
+#endif
+
+#ifdef _MSC_VER
+#define CV_SUPPRESS_DEPRECATED_START \
+    CV_DO_PRAGMA(warning(push)) \
+    CV_DO_PRAGMA(warning(disable: 4996))
+#define CV_SUPPRESS_DEPRECATED_END CV_DO_PRAGMA(warning(pop))
+#elif defined (__clang__) || ((__GNUC__)  && (__GNUC__*100 + __GNUC_MINOR__ > 405))
+#define CV_SUPPRESS_DEPRECATED_START \
+    CV_DO_PRAGMA(GCC diagnostic push) \
+    CV_DO_PRAGMA(GCC diagnostic ignored "-Wdeprecated-declarations")
+#define CV_SUPPRESS_DEPRECATED_END CV_DO_PRAGMA(GCC diagnostic pop)
+#else
+#define CV_SUPPRESS_DEPRECATED_START
+#define CV_SUPPRESS_DEPRECATED_END
+#endif
+
+#define CV_UNUSED(name) (void)name
+
+//! @endcond
+
+// undef problematic defines sometimes defined by system headers (windows.h in particular)
+#undef small
+#undef min
+#undef max
+#undef abs
+#undef Complex
+
+#if defined __cplusplus
+#include <limits>
+#else
+#include <limits.h>
+#endif
+
+#include "opencv2/core/hal/interface.h"
+
+#if defined __ICL
+#  define CV_ICC   __ICL
+#elif defined __ICC
+#  define CV_ICC   __ICC
+#elif defined __ECL
+#  define CV_ICC   __ECL
+#elif defined __ECC
+#  define CV_ICC   __ECC
+#elif defined __INTEL_COMPILER
+#  define CV_ICC   __INTEL_COMPILER
+#endif
+
+#ifndef CV_INLINE
+#  if defined __cplusplus
+#    define CV_INLINE static inline
+#  elif defined _MSC_VER
+#    define CV_INLINE __inline
+#  else
+#    define CV_INLINE static
+#  endif
+#endif
+
+#ifndef CV_ALWAYS_INLINE
+#if defined(__GNUC__) && (__GNUC__ > 3 || (__GNUC__ == 3 && __GNUC_MINOR__ >= 1))
+#define CV_ALWAYS_INLINE inline __attribute__((always_inline))
+#elif defined(_MSC_VER)
+#define CV_ALWAYS_INLINE __forceinline
+#else
+#define CV_ALWAYS_INLINE inline
+#endif
+#endif
+
+#if defined CV_DISABLE_OPTIMIZATION || (defined CV_ICC && !defined CV_ENABLE_UNROLLED)
+#  define CV_ENABLE_UNROLLED 0
+#else
+#  define CV_ENABLE_UNROLLED 1
+#endif
+
+#ifdef __GNUC__
+#  define CV_DECL_ALIGNED(x) __attribute__ ((aligned (x)))
+#elif defined _MSC_VER
+#  define CV_DECL_ALIGNED(x) __declspec(align(x))
+#else
+#  define CV_DECL_ALIGNED(x)
+#endif
+
+/* CPU features and intrinsics support */
+#define CV_CPU_NONE             0
+#define CV_CPU_MMX              1
+#define CV_CPU_SSE              2
+#define CV_CPU_SSE2             3
+#define CV_CPU_SSE3             4
+#define CV_CPU_SSSE3            5
+#define CV_CPU_SSE4_1           6
+#define CV_CPU_SSE4_2           7
+#define CV_CPU_POPCNT           8
+#define CV_CPU_FP16             9
+#define CV_CPU_AVX              10
+#define CV_CPU_AVX2             11
+#define CV_CPU_FMA3             12
+
+#define CV_CPU_AVX_512F         13
+#define CV_CPU_AVX_512BW        14
+#define CV_CPU_AVX_512CD        15
+#define CV_CPU_AVX_512DQ        16
+#define CV_CPU_AVX_512ER        17
+#define CV_CPU_AVX_512IFMA512   18 // deprecated
+#define CV_CPU_AVX_512IFMA      18
+#define CV_CPU_AVX_512PF        19
+#define CV_CPU_AVX_512VBMI      20
+#define CV_CPU_AVX_512VL        21
+#define CV_CPU_AVX_512VBMI2     22
+#define CV_CPU_AVX_512VNNI      23
+#define CV_CPU_AVX_512BITALG    24
+#define CV_CPU_AVX_512VPOPCNTDQ 25
+#define CV_CPU_AVX_5124VNNIW    26
+#define CV_CPU_AVX_5124FMAPS    27
+
+#define CV_CPU_NEON             100
+
+#define CV_CPU_MSA              150
+
+#define CV_CPU_VSX              200
+#define CV_CPU_VSX3             201
+
+// CPU features groups
+#define CV_CPU_AVX512_SKX       256
+#define CV_CPU_AVX512_COMMON    257
+#define CV_CPU_AVX512_KNL       258
+#define CV_CPU_AVX512_KNM       259
+#define CV_CPU_AVX512_CNL       260
+#define CV_CPU_AVX512_CLX       261
+#define CV_CPU_AVX512_ICL       262
+
+// when adding to this list remember to update the following enum
+#define CV_HARDWARE_MAX_FEATURE 512
+
+/** @brief Available CPU features.
+*/
+enum CpuFeatures {
+    CPU_MMX             = 1,
+    CPU_SSE             = 2,
+    CPU_SSE2            = 3,
+    CPU_SSE3            = 4,
+    CPU_SSSE3           = 5,
+    CPU_SSE4_1          = 6,
+    CPU_SSE4_2          = 7,
+    CPU_POPCNT          = 8,
+    CPU_FP16            = 9,
+    CPU_AVX             = 10,
+    CPU_AVX2            = 11,
+    CPU_FMA3            = 12,
+
+    CPU_AVX_512F        = 13,
+    CPU_AVX_512BW       = 14,
+    CPU_AVX_512CD       = 15,
+    CPU_AVX_512DQ       = 16,
+    CPU_AVX_512ER       = 17,
+    CPU_AVX_512IFMA512  = 18, // deprecated
+    CPU_AVX_512IFMA     = 18,
+    CPU_AVX_512PF       = 19,
+    CPU_AVX_512VBMI     = 20,
+    CPU_AVX_512VL       = 21,
+    CPU_AVX_512VBMI2    = 22,
+    CPU_AVX_512VNNI     = 23,
+    CPU_AVX_512BITALG   = 24,
+    CPU_AVX_512VPOPCNTDQ= 25,
+    CPU_AVX_5124VNNIW   = 26,
+    CPU_AVX_5124FMAPS   = 27,
+
+    CPU_NEON            = 100,
+
+    CPU_MSA             = 150,
+
+    CPU_VSX             = 200,
+    CPU_VSX3            = 201,
+
+    CPU_AVX512_SKX      = 256, //!< Skylake-X with AVX-512F/CD/BW/DQ/VL
+    CPU_AVX512_COMMON   = 257, //!< Common instructions AVX-512F/CD for all CPUs that support AVX-512
+    CPU_AVX512_KNL      = 258, //!< Knights Landing with AVX-512F/CD/ER/PF
+    CPU_AVX512_KNM      = 259, //!< Knights Mill with AVX-512F/CD/ER/PF/4FMAPS/4VNNIW/VPOPCNTDQ
+    CPU_AVX512_CNL      = 260, //!< Cannon Lake with AVX-512F/CD/BW/DQ/VL/IFMA/VBMI
+    CPU_AVX512_CLX      = 261, //!< Cascade Lake with AVX-512F/CD/BW/DQ/VL/VNNI
+    CPU_AVX512_ICL      = 262, //!< Ice Lake with AVX-512F/CD/BW/DQ/VL/IFMA/VBMI/VNNI/VBMI2/BITALG/VPOPCNTDQ
+
+    CPU_MAX_FEATURE     = 512  // see CV_HARDWARE_MAX_FEATURE
+};
+
+
+#include "cv_cpu_dispatch.h"
+
+#if !defined(CV_STRONG_ALIGNMENT) && defined(__arm__) && !(defined(__aarch64__) || defined(_M_ARM64))
+// int*, int64* should be propertly aligned pointers on ARMv7
+#define CV_STRONG_ALIGNMENT 1
+#endif
+#if !defined(CV_STRONG_ALIGNMENT)
+#define CV_STRONG_ALIGNMENT 0
+#endif
+
+/* fundamental constants */
+#define CV_PI   3.1415926535897932384626433832795
+#define CV_2PI  6.283185307179586476925286766559
+#define CV_LOG2 0.69314718055994530941723212145818
+
+#if defined __ARM_FP16_FORMAT_IEEE \
+    && !defined __CUDACC__
+#  define CV_FP16_TYPE 1
+#else
+#  define CV_FP16_TYPE 0
+#endif
+
+typedef union Cv16suf
+{
+    short i;
+    ushort u;
+#if CV_FP16_TYPE
+    __fp16 h;
+#endif
+}
+Cv16suf;
+
+typedef union Cv32suf
+{
+    int i;
+    unsigned u;
+    float f;
+}
+Cv32suf;
+
+typedef union Cv64suf
+{
+    int64 i;
+    uint64 u;
+    double f;
+}
+Cv64suf;
+
+#define OPENCV_ABI_COMPATIBILITY 400
+
+#ifdef __OPENCV_BUILD
+#  define DISABLE_OPENCV_3_COMPATIBILITY
+#  define OPENCV_DISABLE_DEPRECATED_COMPATIBILITY
+#endif
+
+#ifndef CV_EXPORTS
+# if (defined _WIN32 || defined WINCE || defined __CYGWIN__) && defined(CVAPI_EXPORTS)
+#   define CV_EXPORTS __declspec(dllexport)
+# elif defined __GNUC__ && __GNUC__ >= 4 && (defined(CVAPI_EXPORTS) || defined(__APPLE__))
+#   define CV_EXPORTS __attribute__ ((visibility ("default")))
+# endif
+#endif
+
+#ifndef CV_EXPORTS
+# define CV_EXPORTS
+#endif
+
+#ifdef _MSC_VER
+#   define CV_EXPORTS_TEMPLATE
+#else
+#   define CV_EXPORTS_TEMPLATE CV_EXPORTS
+#endif
+
+#ifndef CV_DEPRECATED
+#  if defined(__GNUC__)
+#    define CV_DEPRECATED __attribute__ ((deprecated))
+#  elif defined(_MSC_VER)
+#    define CV_DEPRECATED __declspec(deprecated)
+#  else
+#    define CV_DEPRECATED
+#  endif
+#endif
+
+#ifndef CV_DEPRECATED_EXTERNAL
+#  if defined(__OPENCV_BUILD)
+#    define CV_DEPRECATED_EXTERNAL /* nothing */
+#  else
+#    define CV_DEPRECATED_EXTERNAL CV_DEPRECATED
+#  endif
+#endif
+
+
+#ifndef CV_EXTERN_C
+#  ifdef __cplusplus
+#    define CV_EXTERN_C extern "C"
+#  else
+#    define CV_EXTERN_C
+#  endif
+#endif
+
+/* special informative macros for wrapper generators */
+#define CV_EXPORTS_W CV_EXPORTS
+#define CV_EXPORTS_W_SIMPLE CV_EXPORTS
+#define CV_EXPORTS_AS(synonym) CV_EXPORTS
+#define CV_EXPORTS_W_MAP CV_EXPORTS
+#define CV_IN_OUT
+#define CV_OUT
+#define CV_PROP
+#define CV_PROP_RW
+#define CV_WRAP
+#define CV_WRAP_AS(synonym)
+#define CV_WRAP_MAPPABLE(mappable)
+#define CV_WRAP_PHANTOM(phantom_header)
+#define CV_WRAP_DEFAULT(val)
+
+/****************************************************************************************\
+*                                  Matrix type (Mat)                                     *
+\****************************************************************************************/
+
+#define CV_MAT_CN_MASK          ((CV_CN_MAX - 1) << CV_CN_SHIFT)
+#define CV_MAT_CN(flags)        ((((flags) & CV_MAT_CN_MASK) >> CV_CN_SHIFT) + 1)
+#define CV_MAT_TYPE_MASK        (CV_DEPTH_MAX*CV_CN_MAX - 1)
+#define CV_MAT_TYPE(flags)      ((flags) & CV_MAT_TYPE_MASK)
+#define CV_MAT_CONT_FLAG_SHIFT  14
+#define CV_MAT_CONT_FLAG        (1 << CV_MAT_CONT_FLAG_SHIFT)
+#define CV_IS_MAT_CONT(flags)   ((flags) & CV_MAT_CONT_FLAG)
+#define CV_IS_CONT_MAT          CV_IS_MAT_CONT
+#define CV_SUBMAT_FLAG_SHIFT    15
+#define CV_SUBMAT_FLAG          (1 << CV_SUBMAT_FLAG_SHIFT)
+#define CV_IS_SUBMAT(flags)     ((flags) & CV_MAT_SUBMAT_FLAG)
+
+/** Size of each channel item,
+   0x28442211 = 0010 1000 0100 0100 0010 0010 0001 0001 ~ array of sizeof(arr_type_elem) */
+#define CV_ELEM_SIZE1(type) ((0x28442211 >> CV_MAT_DEPTH(type)*4) & 15)
+
+#define CV_ELEM_SIZE(type) (CV_MAT_CN(type)*CV_ELEM_SIZE1(type))
+
+#ifndef MIN
+#  define MIN(a,b)  ((a) > (b) ? (b) : (a))
+#endif
+
+#ifndef MAX
+#  define MAX(a,b)  ((a) < (b) ? (b) : (a))
+#endif
+
+///////////////////////////////////////// Enum operators ///////////////////////////////////////
+
+/**
+
+Provides compatibility operators for both classical and C++11 enum classes,
+as well as exposing the C++11 enum class members for backwards compatibility
+
+@code
+    // Provides operators required for flag enums
+    CV_ENUM_FLAGS(AccessFlag)
+
+    // Exposes the listed members of the enum class AccessFlag to the current namespace
+    CV_ENUM_CLASS_EXPOSE(AccessFlag, ACCESS_READ [, ACCESS_WRITE [, ...] ]);
+@endcode
+*/
+
+#define __CV_ENUM_CLASS_EXPOSE_1(EnumType, MEMBER_CONST)                                              \
+static const EnumType MEMBER_CONST = EnumType::MEMBER_CONST;                                          \
+
+#define __CV_ENUM_CLASS_EXPOSE_2(EnumType, MEMBER_CONST, ...)                                         \
+__CV_ENUM_CLASS_EXPOSE_1(EnumType, MEMBER_CONST);                                                     \
+__CV_EXPAND(__CV_ENUM_CLASS_EXPOSE_1(EnumType, __VA_ARGS__));                                         \
+
+#define __CV_ENUM_CLASS_EXPOSE_3(EnumType, MEMBER_CONST, ...)                                         \
+__CV_ENUM_CLASS_EXPOSE_1(EnumType, MEMBER_CONST);                                                     \
+__CV_EXPAND(__CV_ENUM_CLASS_EXPOSE_2(EnumType, __VA_ARGS__));                                         \
+
+#define __CV_ENUM_CLASS_EXPOSE_4(EnumType, MEMBER_CONST, ...)                                         \
+__CV_ENUM_CLASS_EXPOSE_1(EnumType, MEMBER_CONST);                                                     \
+__CV_EXPAND(__CV_ENUM_CLASS_EXPOSE_3(EnumType, __VA_ARGS__));                                         \
+
+#define __CV_ENUM_CLASS_EXPOSE_5(EnumType, MEMBER_CONST, ...)                                         \
+__CV_ENUM_CLASS_EXPOSE_1(EnumType, MEMBER_CONST);                                                     \
+__CV_EXPAND(__CV_ENUM_CLASS_EXPOSE_4(EnumType, __VA_ARGS__));                                         \
+
+#define __CV_ENUM_CLASS_EXPOSE_6(EnumType, MEMBER_CONST, ...)                                         \
+__CV_ENUM_CLASS_EXPOSE_1(EnumType, MEMBER_CONST);                                                     \
+__CV_EXPAND(__CV_ENUM_CLASS_EXPOSE_5(EnumType, __VA_ARGS__));                                         \
+
+#define __CV_ENUM_CLASS_EXPOSE_7(EnumType, MEMBER_CONST, ...)                                         \
+__CV_ENUM_CLASS_EXPOSE_1(EnumType, MEMBER_CONST);                                                     \
+__CV_EXPAND(__CV_ENUM_CLASS_EXPOSE_6(EnumType, __VA_ARGS__));                                         \
+
+#define __CV_ENUM_CLASS_EXPOSE_8(EnumType, MEMBER_CONST, ...)                                         \
+__CV_ENUM_CLASS_EXPOSE_1(EnumType, MEMBER_CONST);                                                     \
+__CV_EXPAND(__CV_ENUM_CLASS_EXPOSE_7(EnumType, __VA_ARGS__));                                         \
+
+#define __CV_ENUM_CLASS_EXPOSE_9(EnumType, MEMBER_CONST, ...)                                         \
+__CV_ENUM_CLASS_EXPOSE_1(EnumType, MEMBER_CONST);                                                     \
+__CV_EXPAND(__CV_ENUM_CLASS_EXPOSE_8(EnumType, __VA_ARGS__));                                         \
+
+#define __CV_ENUM_FLAGS_LOGICAL_NOT(EnumType)                                                         \
+static inline bool operator!(const EnumType& val)                                                     \
+{                                                                                                     \
+    typedef std::underlying_type<EnumType>::type UnderlyingType;                                      \
+    return !static_cast<UnderlyingType>(val);                                                         \
+}                                                                                                     \
+
+#define __CV_ENUM_FLAGS_LOGICAL_NOT_EQ(Arg1Type, Arg2Type)                                            \
+static inline bool operator!=(const Arg1Type& a, const Arg2Type& b)                                   \
+{                                                                                                     \
+    return static_cast<int>(a) != static_cast<int>(b);                                                \
+}                                                                                                     \
+
+#define __CV_ENUM_FLAGS_LOGICAL_EQ(Arg1Type, Arg2Type)                                                \
+static inline bool operator==(const Arg1Type& a, const Arg2Type& b)                                   \
+{                                                                                                     \
+    return static_cast<int>(a) == static_cast<int>(b);                                                \
+}                                                                                                     \
+
+#define __CV_ENUM_FLAGS_BITWISE_NOT(EnumType)                                                         \
+static inline EnumType operator~(const EnumType& val)                                                 \
+{                                                                                                     \
+    typedef std::underlying_type<EnumType>::type UnderlyingType;                                      \
+    return static_cast<EnumType>(~static_cast<UnderlyingType>(val));                                  \
+}                                                                                                     \
+
+#define __CV_ENUM_FLAGS_BITWISE_OR(EnumType, Arg1Type, Arg2Type)                                      \
+static inline EnumType operator|(const Arg1Type& a, const Arg2Type& b)                                \
+{                                                                                                     \
+    typedef std::underlying_type<EnumType>::type UnderlyingType;                                      \
+    return static_cast<EnumType>(static_cast<UnderlyingType>(a) | static_cast<UnderlyingType>(b));    \
+}                                                                                                     \
+
+#define __CV_ENUM_FLAGS_BITWISE_AND(EnumType, Arg1Type, Arg2Type)                                     \
+static inline EnumType operator&(const Arg1Type& a, const Arg2Type& b)                                \
+{                                                                                                     \
+    typedef std::underlying_type<EnumType>::type UnderlyingType;                                      \
+    return static_cast<EnumType>(static_cast<UnderlyingType>(a) & static_cast<UnderlyingType>(b));    \
+}                                                                                                     \
+
+#define __CV_ENUM_FLAGS_BITWISE_XOR(EnumType, Arg1Type, Arg2Type)                                     \
+static inline EnumType operator^(const Arg1Type& a, const Arg2Type& b)                                \
+{                                                                                                     \
+    typedef std::underlying_type<EnumType>::type UnderlyingType;                                      \
+    return static_cast<EnumType>(static_cast<UnderlyingType>(a) ^ static_cast<UnderlyingType>(b));    \
+}                                                                                                     \
+
+#define __CV_ENUM_FLAGS_BITWISE_OR_EQ(EnumType, Arg1Type)                                             \
+static inline EnumType& operator|=(EnumType& _this, const Arg1Type& val)                              \
+{                                                                                                     \
+    _this = static_cast<EnumType>(static_cast<int>(_this) | static_cast<int>(val));                   \
+    return _this;                                                                                     \
+}                                                                                                     \
+
+#define __CV_ENUM_FLAGS_BITWISE_AND_EQ(EnumType, Arg1Type)                                            \
+static inline EnumType& operator&=(EnumType& _this, const Arg1Type& val)                              \
+{                                                                                                     \
+    _this = static_cast<EnumType>(static_cast<int>(_this) & static_cast<int>(val));                   \
+    return _this;                                                                                     \
+}                                                                                                     \
+
+#define __CV_ENUM_FLAGS_BITWISE_XOR_EQ(EnumType, Arg1Type)                                            \
+static inline EnumType& operator^=(EnumType& _this, const Arg1Type& val)                              \
+{                                                                                                     \
+    _this = static_cast<EnumType>(static_cast<int>(_this) ^ static_cast<int>(val));                   \
+    return _this;                                                                                     \
+}                                                                                                     \
+
+#define CV_ENUM_CLASS_EXPOSE(EnumType, ...)                                                           \
+__CV_EXPAND(__CV_CAT(__CV_ENUM_CLASS_EXPOSE_, __CV_VA_NUM_ARGS(__VA_ARGS__))(EnumType, __VA_ARGS__)); \
+
+#define CV_ENUM_FLAGS(EnumType)                                                                       \
+__CV_ENUM_FLAGS_LOGICAL_NOT      (EnumType)                                                           \
+__CV_ENUM_FLAGS_LOGICAL_EQ       (EnumType, int)                                                      \
+__CV_ENUM_FLAGS_LOGICAL_NOT_EQ   (EnumType, int)                                                      \
+                                                                                                      \
+__CV_ENUM_FLAGS_BITWISE_NOT      (EnumType)                                                           \
+__CV_ENUM_FLAGS_BITWISE_OR       (EnumType, EnumType, EnumType)                                       \
+__CV_ENUM_FLAGS_BITWISE_AND      (EnumType, EnumType, EnumType)                                       \
+__CV_ENUM_FLAGS_BITWISE_XOR      (EnumType, EnumType, EnumType)                                       \
+                                                                                                      \
+__CV_ENUM_FLAGS_BITWISE_OR_EQ    (EnumType, EnumType)                                                 \
+__CV_ENUM_FLAGS_BITWISE_AND_EQ   (EnumType, EnumType)                                                 \
+__CV_ENUM_FLAGS_BITWISE_XOR_EQ   (EnumType, EnumType)                                                 \
+
+/****************************************************************************************\
+*                                    static analysys                                     *
+\****************************************************************************************/
+
+// In practice, some macro are not processed correctly (noreturn is not detected).
+// We need to use simplified definition for them.
+#ifndef CV_STATIC_ANALYSIS
+# if defined(__KLOCWORK__) || defined(__clang_analyzer__) || defined(__COVERITY__)
+#   define CV_STATIC_ANALYSIS 1
+# endif
+#else
+# if defined(CV_STATIC_ANALYSIS) && !(__CV_CAT(1, CV_STATIC_ANALYSIS) == 1)  // defined and not empty
+#   if 0 == CV_STATIC_ANALYSIS
+#     undef CV_STATIC_ANALYSIS
+#   endif
+# endif
+#endif
+
+/****************************************************************************************\
+*                                    Thread sanitizer                                    *
+\****************************************************************************************/
+#ifndef CV_THREAD_SANITIZER
+# if defined(__has_feature)
+#   if __has_feature(thread_sanitizer)
+#     define CV_THREAD_SANITIZER
+#   endif
+# endif
+#endif
+
+/****************************************************************************************\
+*          exchange-add operation for atomic operations on reference counters            *
+\****************************************************************************************/
+
+#ifdef CV_XADD
+  // allow to use user-defined macro
+#elif defined __GNUC__ || defined __clang__
+#  if defined __clang__ && __clang_major__ >= 3 && !defined __ANDROID__ && !defined __EMSCRIPTEN__ && !defined(__CUDACC__)  && !defined __INTEL_COMPILER
+#    ifdef __ATOMIC_ACQ_REL
+#      define CV_XADD(addr, delta) __c11_atomic_fetch_add((_Atomic(int)*)(addr), delta, __ATOMIC_ACQ_REL)
+#    else
+#      define CV_XADD(addr, delta) __atomic_fetch_add((_Atomic(int)*)(addr), delta, 4)
+#    endif
+#  else
+#    if defined __ATOMIC_ACQ_REL && !defined __clang__
+       // version for gcc >= 4.7
+#      define CV_XADD(addr, delta) (int)__atomic_fetch_add((unsigned*)(addr), (unsigned)(delta), __ATOMIC_ACQ_REL)
+#    else
+#      define CV_XADD(addr, delta) (int)__sync_fetch_and_add((unsigned*)(addr), (unsigned)(delta))
+#    endif
+#  endif
+#elif defined _MSC_VER && !defined RC_INVOKED
+#  include <intrin.h>
+#  define CV_XADD(addr, delta) (int)_InterlockedExchangeAdd((long volatile*)addr, delta)
+#else
+  #ifdef OPENCV_FORCE_UNSAFE_XADD
+    CV_INLINE CV_XADD(int* addr, int delta) { int tmp = *addr; *addr += delta; return tmp; }
+  #else
+    #error "OpenCV: can't define safe CV_XADD macro for current platform (unsupported). Define CV_XADD macro through custom port header (see OPENCV_INCLUDE_PORT_FILE)"
+  #endif
+#endif
+
+
+/****************************************************************************************\
+*                                  CV_NORETURN attribute                                 *
+\****************************************************************************************/
+
+#ifndef CV_NORETURN
+#  if defined(__GNUC__)
+#    define CV_NORETURN __attribute__((__noreturn__))
+#  elif defined(_MSC_VER) && (_MSC_VER >= 1300)
+#    define CV_NORETURN __declspec(noreturn)
+#  else
+#    define CV_NORETURN /* nothing by default */
+#  endif
+#endif
+
+
+/****************************************************************************************\
+*                                  CV_NODISCARD attribute                                *
+* encourages the compiler to issue a warning if the return value is discarded (C++17)    *
+\****************************************************************************************/
+#ifndef CV_NODISCARD
+#  if defined(__GNUC__)
+#    define CV_NODISCARD __attribute__((__warn_unused_result__)) // at least available with GCC 3.4
+#  elif defined(__clang__) && defined(__has_attribute)
+#    if __has_attribute(__warn_unused_result__)
+#      define CV_NODISCARD __attribute__((__warn_unused_result__))
+#    endif
+#  endif
+#endif
+#ifndef CV_NODISCARD
+#  define CV_NODISCARD /* nothing by default */
+#endif
+
+
+/****************************************************************************************\
+*                                    C++ 11                                              *
+\****************************************************************************************/
+#ifndef CV_CXX11
+#  if __cplusplus >= 201103L || (defined(_MSC_VER) && _MSC_VER >= 1800)
+#    define CV_CXX11 1
+#  endif
+#else
+#  if CV_CXX11 == 0
+#    undef CV_CXX11
+#  endif
+#endif
+#ifndef CV_CXX11
+#  error "OpenCV 4.x+ requires enabled C++11 support"
+#endif
+
+#define CV_CXX_MOVE_SEMANTICS 1
+#define CV_CXX_MOVE(x) std::move(x)
+#define CV_CXX_STD_ARRAY 1
+#include <array>
+#ifndef CV_OVERRIDE
+#  define CV_OVERRIDE override
+#endif
+#ifndef CV_FINAL
+#  define CV_FINAL final
+#endif
+
+#ifndef CV_NOEXCEPT
+#  if __cplusplus >= 201103L || (defined(_MSC_VER) && _MSC_VER >= 1900/*MSVS 2015*/)
+#    define CV_NOEXCEPT noexcept
+#  endif
+#endif
+#ifndef CV_NOEXCEPT
+#  define CV_NOEXCEPT
+#endif
+
+#ifndef CV_CONSTEXPR
+#  if __cplusplus >= 201103L || (defined(_MSC_VER) && _MSC_VER >= 1900/*MSVS 2015*/)
+#    define CV_CONSTEXPR constexpr
+#  endif
+#endif
+#ifndef CV_CONSTEXPR
+#  define CV_CONSTEXPR
+#endif
+
+// Integer types portatibility
+#ifdef OPENCV_STDINT_HEADER
+#include OPENCV_STDINT_HEADER
+#elif defined(__cplusplus)
+#if defined(_MSC_VER) && _MSC_VER < 1600 /* MSVS 2010 */
+namespace cv {
+typedef signed char int8_t;
+typedef unsigned char uint8_t;
+typedef signed short int16_t;
+typedef unsigned short uint16_t;
+typedef signed int int32_t;
+typedef unsigned int uint32_t;
+typedef signed __int64 int64_t;
+typedef unsigned __int64 uint64_t;
+}
+#elif defined(_MSC_VER) || __cplusplus >= 201103L
+#include <cstdint>
+namespace cv {
+using std::int8_t;
+using std::uint8_t;
+using std::int16_t;
+using std::uint16_t;
+using std::int32_t;
+using std::uint32_t;
+using std::int64_t;
+using std::uint64_t;
+}
+#else
+#include <stdint.h>
+namespace cv {
+typedef ::int8_t int8_t;
+typedef ::uint8_t uint8_t;
+typedef ::int16_t int16_t;
+typedef ::uint16_t uint16_t;
+typedef ::int32_t int32_t;
+typedef ::uint32_t uint32_t;
+typedef ::int64_t int64_t;
+typedef ::uint64_t uint64_t;
+}
+#endif
+#else // pure C
+#include <stdint.h>
+#endif
+
+#ifdef __cplusplus
+namespace cv
+{
+
+class float16_t
+{
+public:
+#if CV_FP16_TYPE
+
+    float16_t() : h(0) {}
+    explicit float16_t(float x) { h = (__fp16)x; }
+    operator float() const { return (float)h; }
+    static float16_t fromBits(ushort w)
+    {
+        Cv16suf u;
+        u.u = w;
+        float16_t result;
+        result.h = u.h;
+        return result;
+    }
+    static float16_t zero()
+    {
+        float16_t result;
+        result.h = (__fp16)0;
+        return result;
+    }
+    ushort bits() const
+    {
+        Cv16suf u;
+        u.h = h;
+        return u.u;
+    }
+protected:
+    __fp16 h;
+
+#else
+    float16_t() : w(0) {}
+    explicit float16_t(float x)
+    {
+    #if CV_AVX2
+        __m128 v = _mm_load_ss(&x);
+        w = (ushort)_mm_cvtsi128_si32(_mm_cvtps_ph(v, 0));
+    #else
+        Cv32suf in;
+        in.f = x;
+        unsigned sign = in.u & 0x80000000;
+        in.u ^= sign;
+
+        if( in.u >= 0x47800000 )
+            w = (ushort)(in.u > 0x7f800000 ? 0x7e00 : 0x7c00);
+        else
+        {
+            if (in.u < 0x38800000)
+            {
+                in.f += 0.5f;
+                w = (ushort)(in.u - 0x3f000000);
+            }
+            else
+            {
+                unsigned t = in.u + 0xc8000fff;
+                w = (ushort)((t + ((in.u >> 13) & 1)) >> 13);
+            }
+        }
+
+        w = (ushort)(w | (sign >> 16));
+    #endif
+    }
+
+    operator float() const
+    {
+    #if CV_AVX2
+        float f;
+        _mm_store_ss(&f, _mm_cvtph_ps(_mm_cvtsi32_si128(w)));
+        return f;
+    #else
+        Cv32suf out;
+
+        unsigned t = ((w & 0x7fff) << 13) + 0x38000000;
+        unsigned sign = (w & 0x8000) << 16;
+        unsigned e = w & 0x7c00;
+
+        out.u = t + (1 << 23);
+        out.u = (e >= 0x7c00 ? t + 0x38000000 :
+                 e == 0 ? (static_cast<void>(out.f -= 6.103515625e-05f), out.u) : t) | sign;
+        return out.f;
+    #endif
+    }
+
+    static float16_t fromBits(ushort b)
+    {
+        float16_t result;
+        result.w = b;
+        return result;
+    }
+    static float16_t zero()
+    {
+        float16_t result;
+        result.w = (ushort)0;
+        return result;
+    }
+    ushort bits() const { return w; }
+protected:
+    ushort w;
+
+#endif
+};
+
+}
+#endif
+
+//! @}
+
+#ifndef __cplusplus
+#include "opencv2/core/fast_math.hpp" // define cvRound(double)
+#endif
+
+#endif // OPENCV_CORE_CVDEF_H

model_zoo/official/lite/app/src/main/cpp/include/opencv2/core/cvstd.hpp

@@ -0,0 +1,190 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CORE_CVSTD_HPP
+#define OPENCV_CORE_CVSTD_HPP
+
+#ifndef __cplusplus
+#  error cvstd.hpp header must be compiled as C++
+#endif
+
+#include "opencv2/core/cvdef.h"
+#include <cstddef>
+#include <cstring>
+#include <cctype>
+
+#include <string>
+
+// import useful primitives from stl
+#  include <algorithm>
+#  include <utility>
+#  include <cstdlib> //for abs(int)
+#  include <cmath>
+
+namespace cv
+{
+    static inline uchar abs(uchar a) { return a; }
+    static inline ushort abs(ushort a) { return a; }
+    static inline unsigned abs(unsigned a) { return a; }
+    static inline uint64 abs(uint64 a) { return a; }
+
+    using std::min;
+    using std::max;
+    using std::abs;
+    using std::swap;
+    using std::sqrt;
+    using std::exp;
+    using std::pow;
+    using std::log;
+}
+
+#include "cvstd_wrapper.hpp"
+
+namespace cv {
+
+//! @addtogroup core_utils
+//! @{
+
+//////////////////////////// memory management functions ////////////////////////////
+
+/** @brief Allocates an aligned memory buffer.
+
+The function allocates the buffer of the specified size and returns it. When the buffer size is 16
+bytes or more, the returned buffer is aligned to 16 bytes.
+@param bufSize Allocated buffer size.
+ */
+CV_EXPORTS void* fastMalloc(size_t bufSize);
+
+/** @brief Deallocates a memory buffer.
+
+The function deallocates the buffer allocated with fastMalloc . If NULL pointer is passed, the
+function does nothing. C version of the function clears the pointer *pptr* to avoid problems with
+double memory deallocation.
+@param ptr Pointer to the allocated buffer.
+ */
+CV_EXPORTS void fastFree(void* ptr);
+
+/*!
+  The STL-compliant memory Allocator based on cv::fastMalloc() and cv::fastFree()
+*/
+template<typename _Tp> class Allocator
+{
+public:
+    typedef _Tp value_type;
+    typedef value_type* pointer;
+    typedef const value_type* const_pointer;
+    typedef value_type& reference;
+    typedef const value_type& const_reference;
+    typedef size_t size_type;
+    typedef ptrdiff_t difference_type;
+    template<typename U> class rebind { typedef Allocator<U> other; };
+
+    explicit Allocator() {}
+    ~Allocator() {}
+    explicit Allocator(Allocator const&) {}
+    template<typename U>
+    explicit Allocator(Allocator<U> const&) {}
+
+    // address
+    pointer address(reference r) { return &r; }
+    const_pointer address(const_reference r) { return &r; }
+
+    pointer allocate(size_type count, const void* =0) { return reinterpret_cast<pointer>(fastMalloc(count * sizeof (_Tp))); }
+    void deallocate(pointer p, size_type) { fastFree(p); }
+
+    void construct(pointer p, const _Tp& v) { new(static_cast<void*>(p)) _Tp(v); }
+    void destroy(pointer p) { p->~_Tp(); }
+
+    size_type max_size() const { return cv::max(static_cast<_Tp>(-1)/sizeof(_Tp), 1); }
+};
+
+//! @} core_utils
+
+//! @endcond
+
+//! @addtogroup core_basic
+//! @{
+
+//////////////////////////////// string class ////////////////////////////////
+
+class CV_EXPORTS FileNode; //for string constructor from FileNode
+
+typedef std::string String;
+
+#ifndef OPENCV_DISABLE_STRING_LOWER_UPPER_CONVERSIONS
+
+//! @cond IGNORED
+namespace details {
+// std::tolower is int->int
+static inline char char_tolower(char ch)
+{
+    return (char)std::tolower((int)ch);
+}
+// std::toupper is int->int
+static inline char char_toupper(char ch)
+{
+    return (char)std::toupper((int)ch);
+}
+} // namespace details
+//! @endcond
+
+static inline std::string toLowerCase(const std::string& str)
+{
+    std::string result(str);
+    std::transform(result.begin(), result.end(), result.begin(), details::char_tolower);
+    return result;
+}
+
+static inline std::string toUpperCase(const std::string& str)
+{
+    std::string result(str);
+    std::transform(result.begin(), result.end(), result.begin(), details::char_toupper);
+    return result;
+}
+
+#endif // OPENCV_DISABLE_STRING_LOWER_UPPER_CONVERSIONS
+
+//! @} core_basic
+} // cv
+
+#endif //OPENCV_CORE_CVSTD_HPP

model_zoo/official/lite/app/src/main/cpp/include/opencv2/core/cvstd.inl.hpp

@@ -0,0 +1,197 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CORE_CVSTDINL_HPP
+#define OPENCV_CORE_CVSTDINL_HPP
+
+#include <complex>
+#include <ostream>
+#include <sstream>
+
+//! @cond IGNORED
+
+#ifdef _MSC_VER
+#pragma warning( push )
+#pragma warning( disable: 4127 )
+#endif
+
+namespace cv
+{
+
+template<typename _Tp> class DataType< std::complex<_Tp> >
+{
+public:
+    typedef std::complex<_Tp>  value_type;
+    typedef value_type         work_type;
+    typedef _Tp                channel_type;
+
+    enum { generic_type = 0,
+           depth        = DataType<channel_type>::depth,
+           channels     = 2,
+           fmt          = DataType<channel_type>::fmt + ((channels - 1) << 8),
+           type         = CV_MAKETYPE(depth, channels) };
+
+    typedef Vec<channel_type, channels> vec_type;
+};
+
+static inline
+std::ostream& operator << (std::ostream& out, Ptr<Formatted> fmtd)
+{
+    fmtd->reset();
+    for(const char* str = fmtd->next(); str; str = fmtd->next())
+        out << str;
+    return out;
+}
+
+static inline
+std::ostream& operator << (std::ostream& out, const Mat& mtx)
+{
+    return out << Formatter::get()->format(mtx);
+}
+
+static inline
+std::ostream& operator << (std::ostream& out, const UMat& m)
+{
+    return out << m.getMat(ACCESS_READ);
+}
+
+template<typename _Tp> static inline
+std::ostream& operator << (std::ostream& out, const Complex<_Tp>& c)
+{
+    return out << "(" << c.re << "," << c.im << ")";
+}
+
+template<typename _Tp> static inline
+std::ostream& operator << (std::ostream& out, const std::vector<Point_<_Tp> >& vec)
+{
+    return out << Formatter::get()->format(Mat(vec));
+}
+
+
+template<typename _Tp> static inline
+std::ostream& operator << (std::ostream& out, const std::vector<Point3_<_Tp> >& vec)
+{
+    return out << Formatter::get()->format(Mat(vec));
+}
+
+
+template<typename _Tp, int m, int n> static inline
+std::ostream& operator << (std::ostream& out, const Matx<_Tp, m, n>& matx)
+{
+    return out << Formatter::get()->format(Mat(matx));
+}
+
+template<typename _Tp> static inline
+std::ostream& operator << (std::ostream& out, const Point_<_Tp>& p)
+{
+    out << "[" << p.x << ", " << p.y << "]";
+    return out;
+}
+
+template<typename _Tp> static inline
+std::ostream& operator << (std::ostream& out, const Point3_<_Tp>& p)
+{
+    out << "[" << p.x << ", " << p.y << ", " << p.z << "]";
+    return out;
+}
+
+template<typename _Tp, int n> static inline
+std::ostream& operator << (std::ostream& out, const Vec<_Tp, n>& vec)
+{
+    out << "[";
+    if (cv::traits::Depth<_Tp>::value <= CV_32S)
+    {
+        for (int i = 0; i < n - 1; ++i) {
+            out << (int)vec[i] << ", ";
+        }
+        out << (int)vec[n-1] << "]";
+    }
+    else
+    {
+        for (int i = 0; i < n - 1; ++i) {
+            out << vec[i] << ", ";
+        }
+        out << vec[n-1] << "]";
+    }
+
+    return out;
+}
+
+template<typename _Tp> static inline
+std::ostream& operator << (std::ostream& out, const Size_<_Tp>& size)
+{
+    return out << "[" << size.width << " x " << size.height << "]";
+}
+
+template<typename _Tp> static inline
+std::ostream& operator << (std::ostream& out, const Rect_<_Tp>& rect)
+{
+    return out << "[" << rect.width << " x " << rect.height << " from (" << rect.x << ", " << rect.y << ")]";
+}
+
+static inline std::ostream& operator << (std::ostream& out, const MatSize& msize)
+{
+    int i, dims = msize.dims();
+    for( i = 0; i < dims; i++ )
+    {
+        out << msize[i];
+        if( i < dims-1 )
+            out << " x ";
+    }
+    return out;
+}
+
+static inline std::ostream &operator<< (std::ostream &s, cv::Range &r)
+{
+    return s << "[" << r.start << " : " << r.end << ")";
+}
+
+} // cv
+
+#ifdef _MSC_VER
+#pragma warning( pop )
+#endif
+
+//! @endcond
+
+#endif // OPENCV_CORE_CVSTDINL_HPP

model_zoo/official/lite/app/src/main/cpp/include/opencv2/core/cvstd_wrapper.hpp

@@ -0,0 +1,154 @@
+// This file is part of OpenCV project.
+// It is subject to the license terms in the LICENSE file found in the top-level directory
+// of this distribution and at http://opencv.org/license.html.
+
+#ifndef OPENCV_CORE_CVSTD_WRAPPER_HPP
+#define OPENCV_CORE_CVSTD_WRAPPER_HPP
+
+#include "opencv2/core/cvdef.h"
+
+#include <string>
+#include <memory>  // std::shared_ptr
+#include <type_traits>  // std::enable_if
+
+namespace cv {
+
+using std::nullptr_t;
+
+//! @addtogroup core_basic
+//! @{
+
+#ifdef CV_DOXYGEN
+
+template <typename _Tp> using Ptr = std::shared_ptr<_Tp>;  // In ideal world it should look like this, but we need some compatibility workarounds below
+
+template<typename _Tp, typename ... A1> static inline
+Ptr<_Tp> makePtr(const A1&... a1) { return std::make_shared<_Tp>(a1...); }
+
+#else  // cv::Ptr with compatibility workarounds
+
+// It should be defined for C-API types only.
+// C++ types should use regular "delete" operator.
+template<typename Y> struct DefaultDeleter;
+#if 0
+{
+    void operator()(Y* p) const;
+};
+#endif
+
+namespace sfinae {
+template<typename C, typename Ret, typename... Args>
+struct has_parenthesis_operator
+{
+private:
+    template<typename T>
+    static CV_CONSTEXPR std::true_type check(typename std::is_same<typename std::decay<decltype(std::declval<T>().operator()(std::declval<Args>()...))>::type, Ret>::type*);
+
+    template<typename> static CV_CONSTEXPR std::false_type check(...);
+
+    typedef decltype(check<C>(0)) type;
+
+public:
+#if __cplusplus >= 201103L || (defined(_MSC_VER) && _MSC_VER >= 1900/*MSVS 2015*/)
+    static CV_CONSTEXPR bool value = type::value;
+#else
+    // support MSVS 2013
+    static const int value = type::value;
+#endif
+};
+} // namespace sfinae
+
+template <typename T, typename = void>
+struct has_custom_delete
+        : public std::false_type {};
+
+// Force has_custom_delete to std::false_type when NVCC is compiling CUDA source files
+#ifndef __CUDACC__
+template <typename T>
+struct has_custom_delete<T, typename std::enable_if< sfinae::has_parenthesis_operator<DefaultDeleter<T>, void, T*>::value >::type >
+        : public std::true_type {};
+#endif
+
+template<typename T>
+struct Ptr : public std::shared_ptr<T>
+{
+#if 0
+    using std::shared_ptr<T>::shared_ptr;  // GCC 5.x can't handle this
+#else
+    inline Ptr() CV_NOEXCEPT : std::shared_ptr<T>() {}
+    inline Ptr(nullptr_t) CV_NOEXCEPT : std::shared_ptr<T>(nullptr) {}
+    template<typename Y, typename D> inline Ptr(Y* p, D d) : std::shared_ptr<T>(p, d) {}
+    template<typename D> inline Ptr(nullptr_t, D d) : std::shared_ptr<T>(nullptr, d) {}
+
+    template<typename Y> inline Ptr(const Ptr<Y>& r, T* ptr) CV_NOEXCEPT : std::shared_ptr<T>(r, ptr) {}
+
+    inline Ptr(const Ptr<T>& o) CV_NOEXCEPT : std::shared_ptr<T>(o) {}
+    inline Ptr(Ptr<T>&& o) CV_NOEXCEPT : std::shared_ptr<T>(std::move(o)) {}
+
+    template<typename Y> inline Ptr(const Ptr<Y>& o) CV_NOEXCEPT : std::shared_ptr<T>(o) {}
+    template<typename Y> inline Ptr(Ptr<Y>&& o) CV_NOEXCEPT : std::shared_ptr<T>(std::move(o)) {}
+#endif
+    inline Ptr(const std::shared_ptr<T>& o) CV_NOEXCEPT : std::shared_ptr<T>(o) {}
+    inline Ptr(std::shared_ptr<T>&& o) CV_NOEXCEPT : std::shared_ptr<T>(std::move(o)) {}
+
+    // Overload with custom DefaultDeleter: Ptr<IplImage>(...)
+    template<typename Y>
+    inline Ptr(const std::true_type&, Y* ptr) : std::shared_ptr<T>(ptr, DefaultDeleter<Y>()) {}
+
+    // Overload without custom deleter: Ptr<std::string>(...);
+    template<typename Y>
+    inline Ptr(const std::false_type&, Y* ptr) : std::shared_ptr<T>(ptr) {}
+
+    template<typename Y = T>
+    inline Ptr(Y* ptr) : Ptr(has_custom_delete<Y>(), ptr) {}
+
+    // Overload with custom DefaultDeleter: Ptr<IplImage>(...)
+    template<typename Y>
+    inline void reset(const std::true_type&, Y* ptr) { std::shared_ptr<T>::reset(ptr, DefaultDeleter<Y>()); }
+
+    // Overload without custom deleter: Ptr<std::string>(...);
+    template<typename Y>
+    inline void reset(const std::false_type&, Y* ptr) { std::shared_ptr<T>::reset(ptr); }
+
+    template<typename Y>
+    inline void reset(Y* ptr) { Ptr<T>::reset(has_custom_delete<Y>(), ptr); }
+
+    template<class Y, class Deleter>
+    void reset(Y* ptr, Deleter d) { std::shared_ptr<T>::reset(ptr, d); }
+
+    void reset() CV_NOEXCEPT { std::shared_ptr<T>::reset(); }
+
+    Ptr& operator=(const Ptr& o) { std::shared_ptr<T>::operator =(o); return *this; }
+    template<typename Y> inline Ptr& operator=(const Ptr<Y>& o) { std::shared_ptr<T>::operator =(o); return *this; }
+
+    T* operator->() const CV_NOEXCEPT { return std::shared_ptr<T>::get();}
+    typename std::add_lvalue_reference<T>::type operator*() const CV_NOEXCEPT { return *std::shared_ptr<T>::get(); }
+
+    // OpenCV 3.x methods (not a part of standard C++ library)
+    inline void release() { std::shared_ptr<T>::reset(); }
+    inline operator T* () const { return std::shared_ptr<T>::get(); }
+    inline bool empty() const { return std::shared_ptr<T>::get() == nullptr; }
+
+    template<typename Y> inline
+    Ptr<Y> staticCast() const CV_NOEXCEPT { return std::static_pointer_cast<Y>(*this); }
+
+    template<typename Y> inline
+    Ptr<Y> constCast() const CV_NOEXCEPT { return std::const_pointer_cast<Y>(*this); }
+
+    template<typename Y> inline
+    Ptr<Y> dynamicCast() const CV_NOEXCEPT { return std::dynamic_pointer_cast<Y>(*this); }
+};
+
+template<typename _Tp, typename ... A1> static inline
+Ptr<_Tp> makePtr(const A1&... a1)
+{
+    static_assert( !has_custom_delete<_Tp>::value, "Can't use this makePtr with custom DefaultDeleter");
+    return (Ptr<_Tp>)std::make_shared<_Tp>(a1...);
+}
+
+#endif // CV_DOXYGEN
+
+//! @} core_basic
+} // cv
+
+#endif //OPENCV_CORE_CVSTD_WRAPPER_HPP

model_zoo/official/lite/app/src/main/cpp/include/opencv2/core/detail/async_promise.hpp

@@ -0,0 +1,71 @@
+// This file is part of OpenCV project.
+// It is subject to the license terms in the LICENSE file found in the top-level directory
+// of this distribution and at http://opencv.org/license.html.
+
+#ifndef OPENCV_CORE_ASYNC_PROMISE_HPP
+#define OPENCV_CORE_ASYNC_PROMISE_HPP
+
+#include "../async.hpp"
+
+#include "exception_ptr.hpp"
+
+namespace cv {
+
+/** @addtogroup core_async
+@{
+*/
+
+
+/** @brief Provides result of asynchronous operations
+
+*/
+class CV_EXPORTS AsyncPromise
+{
+public:
+    ~AsyncPromise() CV_NOEXCEPT;
+    AsyncPromise() CV_NOEXCEPT;
+    explicit AsyncPromise(const AsyncPromise& o) CV_NOEXCEPT;
+    AsyncPromise& operator=(const AsyncPromise& o) CV_NOEXCEPT;
+    void release() CV_NOEXCEPT;
+
+    /** Returns associated AsyncArray
+    @note Can be called once
+    */
+    AsyncArray getArrayResult();
+
+    /** Stores asynchronous result.
+    @param[in] value result
+    */
+    void setValue(InputArray value);
+
+    // TODO "move" setters
+
+#if CV__EXCEPTION_PTR
+    /** Stores exception.
+    @param[in] exception exception to be raised in AsyncArray
+    */
+    void setException(std::exception_ptr exception);
+#endif
+
+    /** Stores exception.
+    @param[in] exception exception to be raised in AsyncArray
+    */
+    void setException(const cv::Exception& exception);
+
+#ifdef CV_CXX11
+    explicit AsyncPromise(AsyncPromise&& o) { p = o.p; o.p = NULL; }
+    AsyncPromise& operator=(AsyncPromise&& o) CV_NOEXCEPT { std::swap(p, o.p); return *this; }
+#endif
+
+
+    // PImpl
+    typedef struct AsyncArray::Impl Impl; friend struct AsyncArray::Impl;
+    inline void* _getImpl() const CV_NOEXCEPT { return p; }
+protected:
+    Impl* p;
+};
+
+
+//! @}
+} // namespace
+#endif // OPENCV_CORE_ASYNC_PROMISE_HPP

model_zoo/official/lite/app/src/main/cpp/include/opencv2/core/detail/exception_ptr.hpp

@@ -0,0 +1,27 @@
+// This file is part of OpenCV project.
+// It is subject to the license terms in the LICENSE file found in the top-level directory
+// of this distribution and at http://opencv.org/license.html.
+
+#ifndef OPENCV_CORE_DETAILS_EXCEPTION_PTR_H
+#define OPENCV_CORE_DETAILS_EXCEPTION_PTR_H
+
+#ifndef CV__EXCEPTION_PTR
+#  if defined(__ANDROID__) && defined(ATOMIC_INT_LOCK_FREE) && ATOMIC_INT_LOCK_FREE < 2
+#    define CV__EXCEPTION_PTR 0  // Not supported, details: https://gcc.gnu.org/bugzilla/show_bug.cgi?id=58938
+#  elif defined(CV_CXX11)
+#    define CV__EXCEPTION_PTR 1
+#  elif defined(_MSC_VER)
+#    define CV__EXCEPTION_PTR (_MSC_VER >= 1600)
+#  elif defined(__clang__)
+#    define CV__EXCEPTION_PTR 0  // C++11 only (see above)
+#  elif defined(__GNUC__) && defined(__GXX_EXPERIMENTAL_CXX0X__)
+#    define CV__EXCEPTION_PTR (__GXX_EXPERIMENTAL_CXX0X__ > 0)
+#  endif
+#endif
+#ifndef CV__EXCEPTION_PTR
+#  define CV__EXCEPTION_PTR 0
+#elif CV__EXCEPTION_PTR
+#  include <exception>  // std::exception_ptr
+#endif
+
+#endif // OPENCV_CORE_DETAILS_EXCEPTION_PTR_H

model_zoo/official/lite/app/src/main/cpp/include/opencv2/core/directx.hpp

@@ -0,0 +1,184 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                           License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2010-2013, Advanced Micro Devices, Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors as is and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the copyright holders or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CORE_DIRECTX_HPP
+#define OPENCV_CORE_DIRECTX_HPP
+
+#include "mat.hpp"
+#include "ocl.hpp"
+
+#if !defined(__d3d11_h__)
+struct ID3D11Device;
+struct ID3D11Texture2D;
+#endif
+
+#if !defined(__d3d10_h__)
+struct ID3D10Device;
+struct ID3D10Texture2D;
+#endif
+
+#if !defined(_D3D9_H_)
+struct IDirect3DDevice9;
+struct IDirect3DDevice9Ex;
+struct IDirect3DSurface9;
+#endif
+
+
+namespace cv { namespace directx {
+
+namespace ocl {
+using namespace cv::ocl;
+
+//! @addtogroup core_directx
+// This section describes OpenCL and DirectX interoperability.
+//
+// To enable DirectX support, configure OpenCV using CMake with WITH_DIRECTX=ON . Note, DirectX is
+// supported only on Windows.
+//
+// To use OpenCL functionality you should first initialize OpenCL context from DirectX resource.
+//
+//! @{
+
+// TODO static functions in the Context class
+//! @brief Creates OpenCL context from D3D11 device
+//
+//! @param pD3D11Device - pointer to D3D11 device
+//! @return Returns reference to OpenCL Context
+CV_EXPORTS Context& initializeContextFromD3D11Device(ID3D11Device* pD3D11Device);
+
+//! @brief Creates OpenCL context from D3D10 device
+//
+//! @param pD3D10Device - pointer to D3D10 device
+//! @return Returns reference to OpenCL Context
+CV_EXPORTS Context& initializeContextFromD3D10Device(ID3D10Device* pD3D10Device);
+
+//! @brief Creates OpenCL context from Direct3DDevice9Ex device
+//
+//! @param pDirect3DDevice9Ex - pointer to Direct3DDevice9Ex device
+//! @return Returns reference to OpenCL Context
+CV_EXPORTS Context& initializeContextFromDirect3DDevice9Ex(IDirect3DDevice9Ex* pDirect3DDevice9Ex);
+
+//! @brief Creates OpenCL context from Direct3DDevice9 device
+//
+//! @param pDirect3DDevice9 - pointer to Direct3Device9 device
+//! @return Returns reference to OpenCL Context
+CV_EXPORTS Context& initializeContextFromDirect3DDevice9(IDirect3DDevice9* pDirect3DDevice9);
+
+//! @}
+
+} // namespace cv::directx::ocl
+
+//! @addtogroup core_directx
+//! @{
+
+//! @brief Converts InputArray to ID3D11Texture2D. If destination texture format is DXGI_FORMAT_NV12 then
+//!        input UMat expected to be in BGR format and data will be downsampled and color-converted to NV12.
+//
+//! @note Note: Destination texture must be allocated by application. Function does memory copy from src to
+//!             pD3D11Texture2D
+//
+//! @param src - source InputArray
+//! @param pD3D11Texture2D - destination D3D11 texture
+CV_EXPORTS void convertToD3D11Texture2D(InputArray src, ID3D11Texture2D* pD3D11Texture2D);
+
+//! @brief Converts ID3D11Texture2D to OutputArray. If input texture format is DXGI_FORMAT_NV12 then
+//!        data will be upsampled and color-converted to BGR format.
+//
+//! @note Note: Destination matrix will be re-allocated if it has not enough memory to match texture size.
+//!             function does memory copy from pD3D11Texture2D to dst
+//
+//! @param pD3D11Texture2D - source D3D11 texture
+//! @param dst             - destination OutputArray
+CV_EXPORTS void convertFromD3D11Texture2D(ID3D11Texture2D* pD3D11Texture2D, OutputArray dst);
+
+//! @brief Converts InputArray to ID3D10Texture2D
+//
+//! @note Note: function does memory copy from src to
+//!             pD3D10Texture2D
+//
+//! @param src             - source InputArray
+//! @param pD3D10Texture2D - destination D3D10 texture
+CV_EXPORTS void convertToD3D10Texture2D(InputArray src, ID3D10Texture2D* pD3D10Texture2D);
+
+//! @brief Converts ID3D10Texture2D to OutputArray
+//
+//! @note Note: function does memory copy from pD3D10Texture2D
+//!             to dst
+//
+//! @param pD3D10Texture2D - source D3D10 texture
+//! @param dst             - destination OutputArray
+CV_EXPORTS void convertFromD3D10Texture2D(ID3D10Texture2D* pD3D10Texture2D, OutputArray dst);
+
+//! @brief Converts InputArray to IDirect3DSurface9
+//
+//! @note Note: function does memory copy from src to
+//!             pDirect3DSurface9
+//
+//! @param src                 - source InputArray
+//! @param pDirect3DSurface9   - destination D3D10 texture
+//! @param surfaceSharedHandle - shared handle
+CV_EXPORTS void convertToDirect3DSurface9(InputArray src, IDirect3DSurface9* pDirect3DSurface9, void* surfaceSharedHandle = NULL);
+
+//! @brief Converts IDirect3DSurface9 to OutputArray
+//
+//! @note Note: function does memory copy from pDirect3DSurface9
+//!             to dst
+//
+//! @param pDirect3DSurface9   - source D3D10 texture
+//! @param dst                 - destination OutputArray
+//! @param surfaceSharedHandle - shared handle
+CV_EXPORTS void convertFromDirect3DSurface9(IDirect3DSurface9* pDirect3DSurface9, OutputArray dst, void* surfaceSharedHandle = NULL);
+
+//! @brief Get OpenCV type from DirectX type
+//! @param iDXGI_FORMAT - enum DXGI_FORMAT for D3D10/D3D11
+//! @return OpenCV type or -1 if there is no equivalent
+CV_EXPORTS int getTypeFromDXGI_FORMAT(const int iDXGI_FORMAT); // enum DXGI_FORMAT for D3D10/D3D11
+
+//! @brief Get OpenCV type from DirectX type
+//! @param iD3DFORMAT - enum D3DTYPE for D3D9
+//! @return OpenCV type or -1 if there is no equivalent
+CV_EXPORTS int getTypeFromD3DFORMAT(const int iD3DFORMAT); // enum D3DTYPE for D3D9
+
+//! @}
+
+} } // namespace cv::directx
+
+#endif // OPENCV_CORE_DIRECTX_HPP

model_zoo/official/lite/app/src/main/cpp/include/opencv2/core/eigen.hpp

@@ -0,0 +1,400 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+
+#ifndef OPENCV_CORE_EIGEN_HPP
+#define OPENCV_CORE_EIGEN_HPP
+
+#ifndef EIGEN_WORLD_VERSION
+#error "Wrong usage of OpenCV's Eigen utility header. Include Eigen's headers first. See https://github.com/opencv/opencv/issues/17366"
+#endif
+
+#include "opencv2/core.hpp"
+
+#if defined _MSC_VER && _MSC_VER >= 1200
+#define NOMINMAX // fix https://github.com/opencv/opencv/issues/17548
+#pragma warning( disable: 4714 ) //__forceinline is not inlined
+#pragma warning( disable: 4127 ) //conditional expression is constant
+#pragma warning( disable: 4244 ) //conversion from '__int64' to 'int', possible loss of data
+#endif
+
+#if EIGEN_WORLD_VERSION == 3 && EIGEN_MAJOR_VERSION >= 3 \
+    && defined(CV_CXX11) && defined(CV_CXX_STD_ARRAY)
+#include <unsupported/Eigen/CXX11/Tensor>
+#define OPENCV_EIGEN_TENSOR_SUPPORT
+#endif // EIGEN_WORLD_VERSION == 3 && EIGEN_MAJOR_VERSION >= 3
+
+namespace cv
+{
+
+/** @addtogroup core_eigen
+These functions are provided for OpenCV-Eigen interoperability. They convert `Mat`
+objects to corresponding `Eigen::Matrix` objects and vice-versa. Consult the [Eigen
+documentation](https://eigen.tuxfamily.org/dox/group__TutorialMatrixClass.html) for
+information about the `Matrix` template type.
+
+@note Using these functions requires the `Eigen/Dense` or similar header to be
+included before this header.
+*/
+//! @{
+
+#if defined(OPENCV_EIGEN_TENSOR_SUPPORT) || defined(CV_DOXYGEN)
+/** @brief Converts an Eigen::Tensor to a cv::Mat.
+
+The method converts an Eigen::Tensor with shape (H x W x C) to a cv::Mat where:
+ H = number of rows
+ W = number of columns
+ C = number of channels
+
+Usage:
+\code
+Eigen::Tensor<float, 3, Eigen::RowMajor> a_tensor(...);
+// populate tensor with values
+Mat a_mat;
+eigen2cv(a_tensor, a_mat);
+\endcode
+*/
+template <typename _Tp, int _layout> static inline
+void eigen2cv( const Eigen::Tensor<_Tp, 3, _layout> &src, OutputArray dst )
+{
+    if( !(_layout & Eigen::RowMajorBit) )
+    {
+        const std::array<int, 3> shuffle{2, 1, 0};
+        Eigen::Tensor<_Tp, 3, !_layout> row_major_tensor = src.swap_layout().shuffle(shuffle);
+        Mat _src(src.dimension(0), src.dimension(1), CV_MAKETYPE(DataType<_Tp>::type, src.dimension(2)), row_major_tensor.data());
+        _src.copyTo(dst);
+    }
+    else
+    {
+        Mat _src(src.dimension(0), src.dimension(1), CV_MAKETYPE(DataType<_Tp>::type, src.dimension(2)), (void *)src.data());
+        _src.copyTo(dst);
+    }
+}
+
+/** @brief Converts a cv::Mat to an Eigen::Tensor.
+
+The method converts a cv::Mat to an Eigen Tensor with shape (H x W x C) where:
+ H = number of rows
+ W = number of columns
+ C = number of channels
+
+Usage:
+\code
+Mat a_mat(...);
+// populate Mat with values
+Eigen::Tensor<float, 3, Eigen::RowMajor> a_tensor(...);
+cv2eigen(a_mat, a_tensor);
+\endcode
+*/
+template <typename _Tp, int _layout> static inline
+void cv2eigen( const Mat &src, Eigen::Tensor<_Tp, 3, _layout> &dst )
+{
+    if( !(_layout & Eigen::RowMajorBit) )
+    {
+        Eigen::Tensor<_Tp, 3, !_layout> row_major_tensor(src.rows, src.cols, src.channels());
+        Mat _dst(src.rows, src.cols, CV_MAKETYPE(DataType<_Tp>::type, src.channels()), row_major_tensor.data());
+        if (src.type() == _dst.type())
+            src.copyTo(_dst);
+        else
+            src.convertTo(_dst, _dst.type());
+        const std::array<int, 3> shuffle{2, 1, 0};
+        dst = row_major_tensor.swap_layout().shuffle(shuffle);
+    }
+    else
+    {
+        dst.resize(src.rows, src.cols, src.channels());
+        Mat _dst(src.rows, src.cols, CV_MAKETYPE(DataType<_Tp>::type, src.channels()), dst.data());
+        if (src.type() == _dst.type())
+            src.copyTo(_dst);
+        else
+            src.convertTo(_dst, _dst.type());
+    }
+}
+
+/** @brief Maps cv::Mat data to an Eigen::TensorMap.
+
+The method wraps an existing Mat data array with an Eigen TensorMap of shape (H x W x C) where:
+ H = number of rows
+ W = number of columns
+ C = number of channels
+
+Explicit instantiation of the return type is required.
+
+@note Caller should be aware of the lifetime of the cv::Mat instance and take appropriate safety measures.
+The cv::Mat instance will retain ownership of the data and the Eigen::TensorMap will lose access when the cv::Mat data is deallocated.
+
+The example below initializes a cv::Mat and produces an Eigen::TensorMap:
+\code
+float arr[] = {0, 1, 2, 3, 4, 5, 6, 7, 8, 9, 10, 11};
+Mat a_mat(2, 2, CV_32FC3, arr);
+Eigen::TensorMap<Eigen::Tensor<float, 3, Eigen::RowMajor>> a_tensormap = cv2eigen_tensormap<float>(a_mat);
+\endcode
+*/
+template <typename _Tp> static inline
+Eigen::TensorMap<Eigen::Tensor<_Tp, 3, Eigen::RowMajor>> cv2eigen_tensormap(InputArray src)
+{
+    Mat mat = src.getMat();
+    CV_CheckTypeEQ(mat.type(), CV_MAKETYPE(traits::Type<_Tp>::value, mat.channels()), "");
+    return Eigen::TensorMap<Eigen::Tensor<_Tp, 3, Eigen::RowMajor>>((_Tp *)mat.data, mat.rows, mat.cols, mat.channels());
+}
+#endif // OPENCV_EIGEN_TENSOR_SUPPORT
+
+template<typename _Tp, int _rows, int _cols, int _options, int _maxRows, int _maxCols> static inline
+void eigen2cv( const Eigen::Matrix<_Tp, _rows, _cols, _options, _maxRows, _maxCols>& src, OutputArray dst )
+{
+    if( !(src.Flags & Eigen::RowMajorBit) )
+    {
+        Mat _src(src.cols(), src.rows(), traits::Type<_Tp>::value,
+              (void*)src.data(), src.outerStride()*sizeof(_Tp));
+        transpose(_src, dst);
+    }
+    else
+    {
+        Mat _src(src.rows(), src.cols(), traits::Type<_Tp>::value,
+                 (void*)src.data(), src.outerStride()*sizeof(_Tp));
+        _src.copyTo(dst);
+    }
+}
+
+// Matx case
+template<typename _Tp, int _rows, int _cols, int _options, int _maxRows, int _maxCols> static inline
+void eigen2cv( const Eigen::Matrix<_Tp, _rows, _cols, _options, _maxRows, _maxCols>& src,
+               Matx<_Tp, _rows, _cols>& dst )
+{
+    if( !(src.Flags & Eigen::RowMajorBit) )
+    {
+        dst = Matx<_Tp, _cols, _rows>(static_cast<const _Tp*>(src.data())).t();
+    }
+    else
+    {
+        dst = Matx<_Tp, _rows, _cols>(static_cast<const _Tp*>(src.data()));
+    }
+}
+
+template<typename _Tp, int _rows, int _cols, int _options, int _maxRows, int _maxCols> static inline
+void cv2eigen( const Mat& src,
+               Eigen::Matrix<_Tp, _rows, _cols, _options, _maxRows, _maxCols>& dst )
+{
+    CV_DbgAssert(src.rows == _rows && src.cols == _cols);
+    if( !(dst.Flags & Eigen::RowMajorBit) )
+    {
+        const Mat _dst(src.cols, src.rows, traits::Type<_Tp>::value,
+                 dst.data(), (size_t)(dst.outerStride()*sizeof(_Tp)));
+        if( src.type() == _dst.type() )
+            transpose(src, _dst);
+        else if( src.cols == src.rows )
+        {
+            src.convertTo(_dst, _dst.type());
+            transpose(_dst, _dst);
+        }
+        else
+            Mat(src.t()).convertTo(_dst, _dst.type());
+    }
+    else
+    {
+        const Mat _dst(src.rows, src.cols, traits::Type<_Tp>::value,
+                 dst.data(), (size_t)(dst.outerStride()*sizeof(_Tp)));
+        src.convertTo(_dst, _dst.type());
+    }
+}
+
+// Matx case
+template<typename _Tp, int _rows, int _cols, int _options, int _maxRows, int _maxCols> static inline
+void cv2eigen( const Matx<_Tp, _rows, _cols>& src,
+               Eigen::Matrix<_Tp, _rows, _cols, _options, _maxRows, _maxCols>& dst )
+{
+    if( !(dst.Flags & Eigen::RowMajorBit) )
+    {
+        const Mat _dst(_cols, _rows, traits::Type<_Tp>::value,
+                 dst.data(), (size_t)(dst.outerStride()*sizeof(_Tp)));
+        transpose(src, _dst);
+    }
+    else
+    {
+        const Mat _dst(_rows, _cols, traits::Type<_Tp>::value,
+                 dst.data(), (size_t)(dst.outerStride()*sizeof(_Tp)));
+        Mat(src).copyTo(_dst);
+    }
+}
+
+template<typename _Tp>  static inline
+void cv2eigen( const Mat& src,
+               Eigen::Matrix<_Tp, Eigen::Dynamic, Eigen::Dynamic>& dst )
+{
+    dst.resize(src.rows, src.cols);
+    if( !(dst.Flags & Eigen::RowMajorBit) )
+    {
+        const Mat _dst(src.cols, src.rows, traits::Type<_Tp>::value,
+             dst.data(), (size_t)(dst.outerStride()*sizeof(_Tp)));
+        if( src.type() == _dst.type() )
+            transpose(src, _dst);
+        else if( src.cols == src.rows )
+        {
+            src.convertTo(_dst, _dst.type());
+            transpose(_dst, _dst);
+        }
+        else
+            Mat(src.t()).convertTo(_dst, _dst.type());
+    }
+    else
+    {
+        const Mat _dst(src.rows, src.cols, traits::Type<_Tp>::value,
+                 dst.data(), (size_t)(dst.outerStride()*sizeof(_Tp)));
+        src.convertTo(_dst, _dst.type());
+    }
+}
+
+// Matx case
+template<typename _Tp, int _rows, int _cols> static inline
+void cv2eigen( const Matx<_Tp, _rows, _cols>& src,
+               Eigen::Matrix<_Tp, Eigen::Dynamic, Eigen::Dynamic>& dst )
+{
+    dst.resize(_rows, _cols);
+    if( !(dst.Flags & Eigen::RowMajorBit) )
+    {
+        const Mat _dst(_cols, _rows, traits::Type<_Tp>::value,
+             dst.data(), (size_t)(dst.outerStride()*sizeof(_Tp)));
+        transpose(src, _dst);
+    }
+    else
+    {
+        const Mat _dst(_rows, _cols, traits::Type<_Tp>::value,
+                 dst.data(), (size_t)(dst.outerStride()*sizeof(_Tp)));
+        Mat(src).copyTo(_dst);
+    }
+}
+
+template<typename _Tp> static inline
+void cv2eigen( const Mat& src,
+               Eigen::Matrix<_Tp, Eigen::Dynamic, 1>& dst )
+{
+    CV_Assert(src.cols == 1);
+    dst.resize(src.rows);
+
+    if( !(dst.Flags & Eigen::RowMajorBit) )
+    {
+        const Mat _dst(src.cols, src.rows, traits::Type<_Tp>::value,
+                 dst.data(), (size_t)(dst.outerStride()*sizeof(_Tp)));
+        if( src.type() == _dst.type() )
+            transpose(src, _dst);
+        else
+            Mat(src.t()).convertTo(_dst, _dst.type());
+    }
+    else
+    {
+        const Mat _dst(src.rows, src.cols, traits::Type<_Tp>::value,
+                 dst.data(), (size_t)(dst.outerStride()*sizeof(_Tp)));
+        src.convertTo(_dst, _dst.type());
+    }
+}
+
+// Matx case
+template<typename _Tp, int _rows> static inline
+void cv2eigen( const Matx<_Tp, _rows, 1>& src,
+               Eigen::Matrix<_Tp, Eigen::Dynamic, 1>& dst )
+{
+    dst.resize(_rows);
+
+    if( !(dst.Flags & Eigen::RowMajorBit) )
+    {
+        const Mat _dst(1, _rows, traits::Type<_Tp>::value,
+                 dst.data(), (size_t)(dst.outerStride()*sizeof(_Tp)));
+        transpose(src, _dst);
+    }
+    else
+    {
+        const Mat _dst(_rows, 1, traits::Type<_Tp>::value,
+                 dst.data(), (size_t)(dst.outerStride()*sizeof(_Tp)));
+        src.copyTo(_dst);
+    }
+}
+
+
+template<typename _Tp> static inline
+void cv2eigen( const Mat& src,
+               Eigen::Matrix<_Tp, 1, Eigen::Dynamic>& dst )
+{
+    CV_Assert(src.rows == 1);
+    dst.resize(src.cols);
+    if( !(dst.Flags & Eigen::RowMajorBit) )
+    {
+        const Mat _dst(src.cols, src.rows, traits::Type<_Tp>::value,
+                 dst.data(), (size_t)(dst.outerStride()*sizeof(_Tp)));
+        if( src.type() == _dst.type() )
+            transpose(src, _dst);
+        else
+            Mat(src.t()).convertTo(_dst, _dst.type());
+    }
+    else
+    {
+        const Mat _dst(src.rows, src.cols, traits::Type<_Tp>::value,
+                 dst.data(), (size_t)(dst.outerStride()*sizeof(_Tp)));
+        src.convertTo(_dst, _dst.type());
+    }
+}
+
+//Matx
+template<typename _Tp, int _cols> static inline
+void cv2eigen( const Matx<_Tp, 1, _cols>& src,
+               Eigen::Matrix<_Tp, 1, Eigen::Dynamic>& dst )
+{
+    dst.resize(_cols);
+    if( !(dst.Flags & Eigen::RowMajorBit) )
+    {
+        const Mat _dst(_cols, 1, traits::Type<_Tp>::value,
+                 dst.data(), (size_t)(dst.outerStride()*sizeof(_Tp)));
+        transpose(src, _dst);
+    }
+    else
+    {
+        const Mat _dst(1, _cols, traits::Type<_Tp>::value,
+                 dst.data(), (size_t)(dst.outerStride()*sizeof(_Tp)));
+        Mat(src).copyTo(_dst);
+    }
+}
+
+//! @}
+
+} // cv
+
+#endif

model_zoo/official/lite/app/src/main/cpp/include/opencv2/core/fast_math.hpp

@@ -0,0 +1,408 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Copyright (C) 2015, Itseez Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_CORE_FAST_MATH_HPP
+#define OPENCV_CORE_FAST_MATH_HPP
+
+#include "opencv2/core/cvdef.h"
+
+//! @addtogroup core_utils
+//! @{
+
+/****************************************************************************************\
+*                                      fast math                                         *
+\****************************************************************************************/
+
+#ifdef __cplusplus
+#  include <cmath>
+#else
+#  ifdef __BORLANDC__
+#    include <fastmath.h>
+#  else
+#    include <math.h>
+#  endif
+#endif
+
+#if defined(__CUDACC__)
+  // nothing, intrinsics/asm code is not supported
+#else
+  #if ((defined _MSC_VER && defined _M_X64) \
+      || (defined __GNUC__ && defined __x86_64__ && defined __SSE2__)) \
+      && !defined(OPENCV_SKIP_INCLUDE_EMMINTRIN_H)
+    #include <emmintrin.h>
+  #endif
+
+  #if defined __PPC64__ && defined __GNUC__ && defined _ARCH_PWR8 \
+      && !defined(OPENCV_SKIP_INCLUDE_ALTIVEC_H)
+    #include <altivec.h>
+  #endif
+
+  #if defined(CV_INLINE_ROUND_FLT)
+    // user-specified version
+    // CV_INLINE_ROUND_DBL should be defined too
+  #elif defined __GNUC__ && defined __arm__ && (defined __ARM_PCS_VFP || defined __ARM_VFPV3__ || defined __ARM_NEON__) && !defined __SOFTFP__
+    // 1. general scheme
+    #define ARM_ROUND(_value, _asm_string) \
+        int res; \
+        float temp; \
+        CV_UNUSED(temp); \
+        __asm__(_asm_string : [res] "=r" (res), [temp] "=w" (temp) : [value] "w" (_value)); \
+        return res
+    // 2. version for double
+    #ifdef __clang__
+        #define CV_INLINE_ROUND_DBL(value) ARM_ROUND(value, "vcvtr.s32.f64 %[temp], %[value] \n vmov %[res], %[temp]")
+    #else
+        #define CV_INLINE_ROUND_DBL(value) ARM_ROUND(value, "vcvtr.s32.f64 %[temp], %P[value] \n vmov %[res], %[temp]")
+    #endif
+    // 3. version for float
+    #define CV_INLINE_ROUND_FLT(value) ARM_ROUND(value, "vcvtr.s32.f32 %[temp], %[value]\n vmov %[res], %[temp]")
+  #elif defined __PPC64__ && defined __GNUC__ && defined _ARCH_PWR8
+    // P8 and newer machines can convert fp32/64 to int quickly.
+    #define CV_INLINE_ROUND_DBL(value) \
+        int out; \
+        double temp; \
+        __asm__( "fctiw %[temp],%[in]\n\tmfvsrwz %[out],%[temp]\n\t" : [out] "=r" (out), [temp] "=d" (temp) : [in] "d" ((double)(value)) : ); \
+        return out;
+
+    // FP32 also works with FP64 routine above
+    #define CV_INLINE_ROUND_FLT(value) CV_INLINE_ROUND_DBL(value)
+  #endif
+
+  #ifdef CV_INLINE_ISINF_FLT
+    // user-specified version
+    // CV_INLINE_ISINF_DBL should be defined too
+  #elif defined __PPC64__ && defined _ARCH_PWR9 && defined(scalar_test_data_class)
+    #define CV_INLINE_ISINF_DBL(value) return scalar_test_data_class(value, 0x30);
+    #define CV_INLINE_ISINF_FLT(value) CV_INLINE_ISINF_DBL(value)
+  #endif
+
+  #ifdef CV_INLINE_ISNAN_FLT
+    // user-specified version
+    // CV_INLINE_ISNAN_DBL should be defined too
+  #elif defined __PPC64__ && defined _ARCH_PWR9 && defined(scalar_test_data_class)
+    #define CV_INLINE_ISNAN_DBL(value) return scalar_test_data_class(value, 0x40);
+    #define CV_INLINE_ISNAN_FLT(value) CV_INLINE_ISNAN_DBL(value)
+  #endif
+
+  #if !defined(OPENCV_USE_FASTMATH_BUILTINS) \
+    && ( \
+        defined(__x86_64__) || defined(__i686__) \
+        || defined(__arm__) \
+        || defined(__PPC64__) \
+    )
+    /* Let builtin C math functions when available. Dedicated hardware is available to
+       round and convert FP values. */
+    #define OPENCV_USE_FASTMATH_BUILTINS 1
+  #endif
+
+  /* Enable builtin math functions if possible, desired, and available.
+     Note, not all math functions inline equally. E.g lrint will not inline
+     without the -fno-math-errno option. */
+  #if defined(CV_ICC)
+    // nothing
+  #elif defined(OPENCV_USE_FASTMATH_BUILTINS) && OPENCV_USE_FASTMATH_BUILTINS
+    #if defined(__clang__)
+      #define CV__FASTMATH_ENABLE_CLANG_MATH_BUILTINS
+      #if !defined(CV_INLINE_ISNAN_DBL) && __has_builtin(__builtin_isnan)
+        #define CV_INLINE_ISNAN_DBL(value) return __builtin_isnan(value);
+      #endif
+      #if !defined(CV_INLINE_ISNAN_FLT) && __has_builtin(__builtin_isnan)
+        #define CV_INLINE_ISNAN_FLT(value) return __builtin_isnan(value);
+      #endif
+      #if !defined(CV_INLINE_ISINF_DBL) && __has_builtin(__builtin_isinf)
+        #define CV_INLINE_ISINF_DBL(value) return __builtin_isinf(value);
+      #endif
+      #if !defined(CV_INLINE_ISINF_FLT) && __has_builtin(__builtin_isinf)
+        #define CV_INLINE_ISINF_FLT(value) return __builtin_isinf(value);
+      #endif
+    #elif defined(__GNUC__)
+      #define CV__FASTMATH_ENABLE_GCC_MATH_BUILTINS
+      #if !defined(CV_INLINE_ISNAN_DBL)
+        #define CV_INLINE_ISNAN_DBL(value) return __builtin_isnan(value);
+      #endif
+      #if !defined(CV_INLINE_ISNAN_FLT)
+        #define CV_INLINE_ISNAN_FLT(value) return __builtin_isnanf(value);
+      #endif
+      #if !defined(CV_INLINE_ISINF_DBL)
+        #define CV_INLINE_ISINF_DBL(value) return __builtin_isinf(value);
+      #endif
+      #if !defined(CV_INLINE_ISINF_FLT)
+        #define CV_INLINE_ISINF_FLT(value) return __builtin_isinff(value);
+      #endif
+    #elif defined(_MSC_VER)
+      #if !defined(CV_INLINE_ISNAN_DBL)
+        #define CV_INLINE_ISNAN_DBL(value) return isnan(value);
+      #endif
+      #if !defined(CV_INLINE_ISNAN_FLT)
+        #define CV_INLINE_ISNAN_FLT(value) return isnan(value);
+      #endif
+      #if !defined(CV_INLINE_ISINF_DBL)
+        #define CV_INLINE_ISINF_DBL(value) return isinf(value);
+      #endif
+      #if !defined(CV_INLINE_ISINF_FLT)
+        #define CV_INLINE_ISINF_FLT(value) return isinf(value);
+      #endif
+    #endif
+  #endif
+
+#endif // defined(__CUDACC__)
+
+/** @brief Rounds floating-point number to the nearest integer
+
+ @param value floating-point number. If the value is outside of INT_MIN ... INT_MAX range, the
+ result is not defined.
+ */
+CV_INLINE int
+cvRound( double value )
+{
+#if defined CV_INLINE_ROUND_DBL
+    CV_INLINE_ROUND_DBL(value);
+#elif ((defined _MSC_VER && defined _M_X64) || (defined __GNUC__ && defined __x86_64__ \
+    && defined __SSE2__ && !defined __APPLE__) || CV_SSE2) \
+    && !defined(__CUDACC__)
+    __m128d t = _mm_set_sd( value );
+    return _mm_cvtsd_si32(t);
+#elif defined _MSC_VER && defined _M_IX86
+    int t;
+    __asm
+    {
+        fld value;
+        fistp t;
+    }
+    return t;
+#elif defined CV_ICC || defined __GNUC__
+    return (int)(lrint(value));
+#else
+    /* it's ok if round does not comply with IEEE754 standard;
+       the tests should allow +/-1 difference when the tested functions use round */
+    return (int)(value + (value >= 0 ? 0.5 : -0.5));
+#endif
+}
+
+
+/** @brief Rounds floating-point number to the nearest integer not larger than the original.
+
+ The function computes an integer i such that:
+ \f[i \le \texttt{value} < i+1\f]
+ @param value floating-point number. If the value is outside of INT_MIN ... INT_MAX range, the
+ result is not defined.
+ */
+CV_INLINE int cvFloor( double value )
+{
+#if (defined CV__FASTMATH_ENABLE_GCC_MATH_BUILTINS || defined CV__FASTMATH_ENABLE_CLANG_MATH_BUILTINS) \
+    && ( \
+        defined(__PPC64__) \
+    )
+    return __builtin_floor(value);
+#else
+    int i = (int)value;
+    return i - (i > value);
+#endif
+}
+
+/** @brief Rounds floating-point number to the nearest integer not smaller than the original.
+
+ The function computes an integer i such that:
+ \f[i \le \texttt{value} < i+1\f]
+ @param value floating-point number. If the value is outside of INT_MIN ... INT_MAX range, the
+ result is not defined.
+ */
+CV_INLINE int cvCeil( double value )
+{
+#if (defined CV__FASTMATH_ENABLE_GCC_MATH_BUILTINS || defined CV__FASTMATH_ENABLE_CLANG_MATH_BUILTINS) \
+    && ( \
+        defined(__PPC64__) \
+    )
+    return __builtin_ceil(value);
+#else
+    int i = (int)value;
+    return i + (i < value);
+#endif
+}
+
+/** @brief Determines if the argument is Not A Number.
+
+ @param value The input floating-point value
+
+ The function returns 1 if the argument is Not A Number (as defined by IEEE754 standard), 0
+ otherwise. */
+CV_INLINE int cvIsNaN( double value )
+{
+#if defined CV_INLINE_ISNAN_DBL
+    CV_INLINE_ISNAN_DBL(value);
+#else
+    Cv64suf ieee754;
+    ieee754.f = value;
+    return ((unsigned)(ieee754.u >> 32) & 0x7fffffff) +
+           ((unsigned)ieee754.u != 0) > 0x7ff00000;
+#endif
+}
+
+/** @brief Determines if the argument is Infinity.
+
+ @param value The input floating-point value
+
+ The function returns 1 if the argument is a plus or minus infinity (as defined by IEEE754 standard)
+ and 0 otherwise. */
+CV_INLINE int cvIsInf( double value )
+{
+#if defined CV_INLINE_ISINF_DBL
+    CV_INLINE_ISINF_DBL(value);
+#elif defined(__x86_64__) || defined(_M_X64) || defined(__aarch64__) || defined(_M_ARM64) || defined(__PPC64__)
+    Cv64suf ieee754;
+    ieee754.f = value;
+    return (ieee754.u & 0x7fffffff00000000) ==
+                        0x7ff0000000000000;
+#else
+    Cv64suf ieee754;
+    ieee754.f = value;
+    return ((unsigned)(ieee754.u >> 32) & 0x7fffffff) == 0x7ff00000 &&
+            (unsigned)ieee754.u == 0;
+#endif
+}
+
+#ifdef __cplusplus
+
+/** @overload */
+CV_INLINE int cvRound(float value)
+{
+#if defined CV_INLINE_ROUND_FLT
+    CV_INLINE_ROUND_FLT(value);
+#elif ((defined _MSC_VER && defined _M_X64) || (defined __GNUC__ && defined __x86_64__ \
+    && defined __SSE2__ && !defined __APPLE__) || CV_SSE2) \
+    && !defined(__CUDACC__)
+    __m128 t = _mm_set_ss( value );
+    return _mm_cvtss_si32(t);
+#elif defined _MSC_VER && defined _M_IX86
+    int t;
+    __asm
+    {
+        fld value;
+        fistp t;
+    }
+    return t;
+#elif defined CV_ICC || defined __GNUC__
+    return (int)(lrintf(value));
+#else
+    /* it's ok if round does not comply with IEEE754 standard;
+     the tests should allow +/-1 difference when the tested functions use round */
+    return (int)(value + (value >= 0 ? 0.5f : -0.5f));
+#endif
+}
+
+/** @overload */
+CV_INLINE int cvRound( int value )
+{
+    return value;
+}
+
+/** @overload */
+CV_INLINE int cvFloor( float value )
+{
+#if (defined CV__FASTMATH_ENABLE_GCC_MATH_BUILTINS || defined CV__FASTMATH_ENABLE_CLANG_MATH_BUILTINS) \
+    && ( \
+        defined(__PPC64__) \
+    )
+    return __builtin_floorf(value);
+#else
+    int i = (int)value;
+    return i - (i > value);
+#endif
+}
+
+/** @overload */
+CV_INLINE int cvFloor( int value )
+{
+    return value;
+}
+
+/** @overload */
+CV_INLINE int cvCeil( float value )
+{
+#if (defined CV__FASTMATH_ENABLE_GCC_MATH_BUILTINS || defined CV__FASTMATH_ENABLE_CLANG_MATH_BUILTINS) \
+    && ( \
+        defined(__PPC64__) \
+    )
+    return __builtin_ceilf(value);
+#else
+    int i = (int)value;
+    return i + (i < value);
+#endif
+}
+
+/** @overload */
+CV_INLINE int cvCeil( int value )
+{
+    return value;
+}
+
+/** @overload */
+CV_INLINE int cvIsNaN( float value )
+{
+#if defined CV_INLINE_ISNAN_FLT
+    CV_INLINE_ISNAN_FLT(value);
+#else
+    Cv32suf ieee754;
+    ieee754.f = value;
+    return (ieee754.u & 0x7fffffff) > 0x7f800000;
+#endif
+}
+
+/** @overload */
+CV_INLINE int cvIsInf( float value )
+{
+#if defined CV_INLINE_ISINF_FLT
+    CV_INLINE_ISINF_FLT(value);
+#else
+    Cv32suf ieee754;
+    ieee754.f = value;
+    return (ieee754.u & 0x7fffffff) == 0x7f800000;
+#endif
+}
+
+#endif // __cplusplus
+
+//! @} core_utils
+
+#endif

model_zoo/official/lite/app/src/main/cpp/include/opencv2/core/hal/hal.hpp

@@ -0,0 +1,256 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Copyright (C) 2015, Itseez Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_HAL_HPP
+#define OPENCV_HAL_HPP
+
+#include "opencv2/core/cvdef.h"
+#include "opencv2/core/cvstd.hpp"
+#include "opencv2/core/hal/interface.h"
+
+namespace cv { namespace hal {
+
+//! @addtogroup core_hal_functions
+//! @{
+
+CV_EXPORTS int normHamming(const uchar* a, int n);
+CV_EXPORTS int normHamming(const uchar* a, const uchar* b, int n);
+
+CV_EXPORTS int normHamming(const uchar* a, int n, int cellSize);
+CV_EXPORTS int normHamming(const uchar* a, const uchar* b, int n, int cellSize);
+
+CV_EXPORTS int LU32f(float* A, size_t astep, int m, float* b, size_t bstep, int n);
+CV_EXPORTS int LU64f(double* A, size_t astep, int m, double* b, size_t bstep, int n);
+CV_EXPORTS bool Cholesky32f(float* A, size_t astep, int m, float* b, size_t bstep, int n);
+CV_EXPORTS bool Cholesky64f(double* A, size_t astep, int m, double* b, size_t bstep, int n);
+CV_EXPORTS void SVD32f(float* At, size_t astep, float* W, float* U, size_t ustep, float* Vt, size_t vstep, int m, int n, int flags);
+CV_EXPORTS void SVD64f(double* At, size_t astep, double* W, double* U, size_t ustep, double* Vt, size_t vstep, int m, int n, int flags);
+CV_EXPORTS int QR32f(float* A, size_t astep, int m, int n, int k, float* b, size_t bstep, float* hFactors);
+CV_EXPORTS int QR64f(double* A, size_t astep, int m, int n, int k, double* b, size_t bstep, double* hFactors);
+
+CV_EXPORTS void gemm32f(const float* src1, size_t src1_step, const float* src2, size_t src2_step,
+                        float alpha, const float* src3, size_t src3_step, float beta, float* dst, size_t dst_step,
+                        int m_a, int n_a, int n_d, int flags);
+CV_EXPORTS void gemm64f(const double* src1, size_t src1_step, const double* src2, size_t src2_step,
+                        double alpha, const double* src3, size_t src3_step, double beta, double* dst, size_t dst_step,
+                        int m_a, int n_a, int n_d, int flags);
+CV_EXPORTS void gemm32fc(const float* src1, size_t src1_step, const float* src2, size_t src2_step,
+                        float alpha, const float* src3, size_t src3_step, float beta, float* dst, size_t dst_step,
+                        int m_a, int n_a, int n_d, int flags);
+CV_EXPORTS void gemm64fc(const double* src1, size_t src1_step, const double* src2, size_t src2_step,
+                        double alpha, const double* src3, size_t src3_step, double beta, double* dst, size_t dst_step,
+                        int m_a, int n_a, int n_d, int flags);
+
+CV_EXPORTS int normL1_(const uchar* a, const uchar* b, int n);
+CV_EXPORTS float normL1_(const float* a, const float* b, int n);
+CV_EXPORTS float normL2Sqr_(const float* a, const float* b, int n);
+
+CV_EXPORTS void exp32f(const float* src, float* dst, int n);
+CV_EXPORTS void exp64f(const double* src, double* dst, int n);
+CV_EXPORTS void log32f(const float* src, float* dst, int n);
+CV_EXPORTS void log64f(const double* src, double* dst, int n);
+
+CV_EXPORTS void fastAtan32f(const float* y, const float* x, float* dst, int n, bool angleInDegrees);
+CV_EXPORTS void fastAtan64f(const double* y, const double* x, double* dst, int n, bool angleInDegrees);
+CV_EXPORTS void magnitude32f(const float* x, const float* y, float* dst, int n);
+CV_EXPORTS void magnitude64f(const double* x, const double* y, double* dst, int n);
+CV_EXPORTS void sqrt32f(const float* src, float* dst, int len);
+CV_EXPORTS void sqrt64f(const double* src, double* dst, int len);
+CV_EXPORTS void invSqrt32f(const float* src, float* dst, int len);
+CV_EXPORTS void invSqrt64f(const double* src, double* dst, int len);
+
+CV_EXPORTS void split8u(const uchar* src, uchar** dst, int len, int cn );
+CV_EXPORTS void split16u(const ushort* src, ushort** dst, int len, int cn );
+CV_EXPORTS void split32s(const int* src, int** dst, int len, int cn );
+CV_EXPORTS void split64s(const int64* src, int64** dst, int len, int cn );
+
+CV_EXPORTS void merge8u(const uchar** src, uchar* dst, int len, int cn );
+CV_EXPORTS void merge16u(const ushort** src, ushort* dst, int len, int cn );
+CV_EXPORTS void merge32s(const int** src, int* dst, int len, int cn );
+CV_EXPORTS void merge64s(const int64** src, int64* dst, int len, int cn );
+
+CV_EXPORTS void add8u( const uchar* src1, size_t step1, const uchar* src2, size_t step2, uchar* dst, size_t step, int width, int height, void* );
+CV_EXPORTS void add8s( const schar* src1, size_t step1, const schar* src2, size_t step2, schar* dst, size_t step, int width, int height, void* );
+CV_EXPORTS void add16u( const ushort* src1, size_t step1, const ushort* src2, size_t step2, ushort* dst, size_t step, int width, int height, void* );
+CV_EXPORTS void add16s( const short* src1, size_t step1, const short* src2, size_t step2, short* dst, size_t step, int width, int height, void* );
+CV_EXPORTS void add32s( const int* src1, size_t step1, const int* src2, size_t step2, int* dst, size_t step, int width, int height, void* );
+CV_EXPORTS void add32f( const float* src1, size_t step1, const float* src2, size_t step2, float* dst, size_t step, int width, int height, void* );
+CV_EXPORTS void add64f( const double* src1, size_t step1, const double* src2, size_t step2, double* dst, size_t step, int width, int height, void* );
+
+CV_EXPORTS void sub8u( const uchar* src1, size_t step1, const uchar* src2, size_t step2, uchar* dst, size_t step, int width, int height, void* );
+CV_EXPORTS void sub8s( const schar* src1, size_t step1, const schar* src2, size_t step2, schar* dst, size_t step, int width, int height, void* );
+CV_EXPORTS void sub16u( const ushort* src1, size_t step1, const ushort* src2, size_t step2, ushort* dst, size_t step, int width, int height, void* );
+CV_EXPORTS void sub16s( const short* src1, size_t step1, const short* src2, size_t step2, short* dst, size_t step, int width, int height, void* );
+CV_EXPORTS void sub32s( const int* src1, size_t step1, const int* src2, size_t step2, int* dst, size_t step, int width, int height, void* );
+CV_EXPORTS void sub32f( const float* src1, size_t step1, const float* src2, size_t step2, float* dst, size_t step, int width, int height, void* );
+CV_EXPORTS void sub64f( const double* src1, size_t step1, const double* src2, size_t step2, double* dst, size_t step, int width, int height, void* );
+
+CV_EXPORTS void max8u( const uchar* src1, size_t step1, const uchar* src2, size_t step2, uchar* dst, size_t step, int width, int height, void* );
+CV_EXPORTS void max8s( const schar* src1, size_t step1, const schar* src2, size_t step2, schar* dst, size_t step, int width, int height, void* );
+CV_EXPORTS void max16u( const ushort* src1, size_t step1, const ushort* src2, size_t step2, ushort* dst, size_t step, int width, int height, void* );
+CV_EXPORTS void max16s( const short* src1, size_t step1, const short* src2, size_t step2, short* dst, size_t step, int width, int height, void* );
+CV_EXPORTS void max32s( const int* src1, size_t step1, const int* src2, size_t step2, int* dst, size_t step, int width, int height, void* );
+CV_EXPORTS void max32f( const float* src1, size_t step1, const float* src2, size_t step2, float* dst, size_t step, int width, int height, void* );
+CV_EXPORTS void max64f( const double* src1, size_t step1, const double* src2, size_t step2, double* dst, size_t step, int width, int height, void* );
+
+CV_EXPORTS void min8u( const uchar* src1, size_t step1, const uchar* src2, size_t step2, uchar* dst, size_t step, int width, int height, void* );
+CV_EXPORTS void min8s( const schar* src1, size_t step1, const schar* src2, size_t step2, schar* dst, size_t step, int width, int height, void* );
+CV_EXPORTS void min16u( const ushort* src1, size_t step1, const ushort* src2, size_t step2, ushort* dst, size_t step, int width, int height, void* );
+CV_EXPORTS void min16s( const short* src1, size_t step1, const short* src2, size_t step2, short* dst, size_t step, int width, int height, void* );
+CV_EXPORTS void min32s( const int* src1, size_t step1, const int* src2, size_t step2, int* dst, size_t step, int width, int height, void* );
+CV_EXPORTS void min32f( const float* src1, size_t step1, const float* src2, size_t step2, float* dst, size_t step, int width, int height, void* );
+CV_EXPORTS void min64f( const double* src1, size_t step1, const double* src2, size_t step2, double* dst, size_t step, int width, int height, void* );
+
+CV_EXPORTS void absdiff8u( const uchar* src1, size_t step1, const uchar* src2, size_t step2, uchar* dst, size_t step, int width, int height, void* );
+CV_EXPORTS void absdiff8s( const schar* src1, size_t step1, const schar* src2, size_t step2, schar* dst, size_t step, int width, int height, void* );
+CV_EXPORTS void absdiff16u( const ushort* src1, size_t step1, const ushort* src2, size_t step2, ushort* dst, size_t step, int width, int height, void* );
+CV_EXPORTS void absdiff16s( const short* src1, size_t step1, const short* src2, size_t step2, short* dst, size_t step, int width, int height, void* );
+CV_EXPORTS void absdiff32s( const int* src1, size_t step1, const int* src2, size_t step2, int* dst, size_t step, int width, int height, void* );
+CV_EXPORTS void absdiff32f( const float* src1, size_t step1, const float* src2, size_t step2, float* dst, size_t step, int width, int height, void* );
+CV_EXPORTS void absdiff64f( const double* src1, size_t step1, const double* src2, size_t step2, double* dst, size_t step, int width, int height, void* );
+
+CV_EXPORTS void and8u( const uchar* src1, size_t step1, const uchar* src2, size_t step2, uchar* dst, size_t step, int width, int height, void* );
+CV_EXPORTS void or8u( const uchar* src1, size_t step1, const uchar* src2, size_t step2, uchar* dst, size_t step, int width, int height, void* );
+CV_EXPORTS void xor8u( const uchar* src1, size_t step1, const uchar* src2, size_t step2, uchar* dst, size_t step, int width, int height, void* );
+CV_EXPORTS void not8u( const uchar* src1, size_t step1, const uchar* src2, size_t step2, uchar* dst, size_t step, int width, int height, void* );
+
+CV_EXPORTS void cmp8u(const uchar* src1, size_t step1, const uchar* src2, size_t step2, uchar* dst, size_t step, int width, int height, void* _cmpop);
+CV_EXPORTS void cmp8s(const schar* src1, size_t step1, const schar* src2, size_t step2, uchar* dst, size_t step, int width, int height, void* _cmpop);
+CV_EXPORTS void cmp16u(const ushort* src1, size_t step1, const ushort* src2, size_t step2, uchar* dst, size_t step, int width, int height, void* _cmpop);
+CV_EXPORTS void cmp16s(const short* src1, size_t step1, const short* src2, size_t step2, uchar* dst, size_t step, int width, int height, void* _cmpop);
+CV_EXPORTS void cmp32s(const int* src1, size_t step1, const int* src2, size_t step2, uchar* dst, size_t step, int width, int height, void* _cmpop);
+CV_EXPORTS void cmp32f(const float* src1, size_t step1, const float* src2, size_t step2, uchar* dst, size_t step, int width, int height, void* _cmpop);
+CV_EXPORTS void cmp64f(const double* src1, size_t step1, const double* src2, size_t step2, uchar* dst, size_t step, int width, int height, void* _cmpop);
+
+CV_EXPORTS void mul8u( const uchar* src1, size_t step1, const uchar* src2, size_t step2, uchar* dst, size_t step, int width, int height, void* scale);
+CV_EXPORTS void mul8s( const schar* src1, size_t step1, const schar* src2, size_t step2, schar* dst, size_t step, int width, int height, void* scale);
+CV_EXPORTS void mul16u( const ushort* src1, size_t step1, const ushort* src2, size_t step2, ushort* dst, size_t step, int width, int height, void* scale);
+CV_EXPORTS void mul16s( const short* src1, size_t step1, const short* src2, size_t step2, short* dst, size_t step, int width, int height, void* scale);
+CV_EXPORTS void mul32s( const int* src1, size_t step1, const int* src2, size_t step2, int* dst, size_t step, int width, int height, void* scale);
+CV_EXPORTS void mul32f( const float* src1, size_t step1, const float* src2, size_t step2, float* dst, size_t step, int width, int height, void* scale);
+CV_EXPORTS void mul64f( const double* src1, size_t step1, const double* src2, size_t step2, double* dst, size_t step, int width, int height, void* scale);
+
+CV_EXPORTS void div8u( const uchar* src1, size_t step1, const uchar* src2, size_t step2, uchar* dst, size_t step, int width, int height, void* scale);
+CV_EXPORTS void div8s( const schar* src1, size_t step1, const schar* src2, size_t step2, schar* dst, size_t step, int width, int height, void* scale);
+CV_EXPORTS void div16u( const ushort* src1, size_t step1, const ushort* src2, size_t step2, ushort* dst, size_t step, int width, int height, void* scale);
+CV_EXPORTS void div16s( const short* src1, size_t step1, const short* src2, size_t step2, short* dst, size_t step, int width, int height, void* scale);
+CV_EXPORTS void div32s( const int* src1, size_t step1, const int* src2, size_t step2, int* dst, size_t step, int width, int height, void* scale);
+CV_EXPORTS void div32f( const float* src1, size_t step1, const float* src2, size_t step2, float* dst, size_t step, int width, int height, void* scale);
+CV_EXPORTS void div64f( const double* src1, size_t step1, const double* src2, size_t step2, double* dst, size_t step, int width, int height, void* scale);
+
+CV_EXPORTS void recip8u( const uchar *, size_t, const uchar * src2, size_t step2, uchar* dst, size_t step, int width, int height, void* scale);
+CV_EXPORTS void recip8s( const schar *, size_t, const schar * src2, size_t step2, schar* dst, size_t step, int width, int height, void* scale);
+CV_EXPORTS void recip16u( const ushort *, size_t, const ushort * src2, size_t step2, ushort* dst, size_t step, int width, int height, void* scale);
+CV_EXPORTS void recip16s( const short *, size_t, const short * src2, size_t step2, short* dst, size_t step, int width, int height, void* scale);
+CV_EXPORTS void recip32s( const int *, size_t, const int * src2, size_t step2, int* dst, size_t step, int width, int height, void* scale);
+CV_EXPORTS void recip32f( const float *, size_t, const float * src2, size_t step2, float* dst, size_t step, int width, int height, void* scale);
+CV_EXPORTS void recip64f( const double *, size_t, const double * src2, size_t step2, double* dst, size_t step, int width, int height, void* scale);
+
+CV_EXPORTS void addWeighted8u( const uchar* src1, size_t step1, const uchar* src2, size_t step2, uchar* dst, size_t step, int width, int height, void* _scalars );
+CV_EXPORTS void addWeighted8s( const schar* src1, size_t step1, const schar* src2, size_t step2, schar* dst, size_t step, int width, int height, void* scalars );
+CV_EXPORTS void addWeighted16u( const ushort* src1, size_t step1, const ushort* src2, size_t step2, ushort* dst, size_t step, int width, int height, void* scalars );
+CV_EXPORTS void addWeighted16s( const short* src1, size_t step1, const short* src2, size_t step2, short* dst, size_t step, int width, int height, void* scalars );
+CV_EXPORTS void addWeighted32s( const int* src1, size_t step1, const int* src2, size_t step2, int* dst, size_t step, int width, int height, void* scalars );
+CV_EXPORTS void addWeighted32f( const float* src1, size_t step1, const float* src2, size_t step2, float* dst, size_t step, int width, int height, void* scalars );
+CV_EXPORTS void addWeighted64f( const double* src1, size_t step1, const double* src2, size_t step2, double* dst, size_t step, int width, int height, void* scalars );
+
+CV_EXPORTS void cvt16f32f( const float16_t* src, float* dst, int len );
+CV_EXPORTS void cvt32f16f( const float* src, float16_t* dst, int len );
+
+CV_EXPORTS void addRNGBias32f( float* arr, const float* scaleBiasPairs, int len );
+CV_EXPORTS void addRNGBias64f( double* arr, const double* scaleBiasPairs, int len );
+
+struct CV_EXPORTS DFT1D
+{
+    static Ptr<DFT1D> create(int len, int count, int depth, int flags, bool * useBuffer = 0);
+    virtual void apply(const uchar *src, uchar *dst) = 0;
+    virtual ~DFT1D() {}
+};
+
+struct CV_EXPORTS DFT2D
+{
+    static Ptr<DFT2D> create(int width, int height, int depth,
+                             int src_channels, int dst_channels,
+                             int flags, int nonzero_rows = 0);
+    virtual void apply(const uchar *src_data, size_t src_step, uchar *dst_data, size_t dst_step) = 0;
+    virtual ~DFT2D() {}
+};
+
+struct CV_EXPORTS DCT2D
+{
+    static Ptr<DCT2D> create(int width, int height, int depth, int flags);
+    virtual void apply(const uchar *src_data, size_t src_step, uchar *dst_data, size_t dst_step) = 0;
+    virtual ~DCT2D() {}
+};
+
+//! @} core_hal
+
+//=============================================================================
+// for binary compatibility with 3.0
+
+//! @cond IGNORED
+
+CV_EXPORTS int LU(float* A, size_t astep, int m, float* b, size_t bstep, int n);
+CV_EXPORTS int LU(double* A, size_t astep, int m, double* b, size_t bstep, int n);
+CV_EXPORTS bool Cholesky(float* A, size_t astep, int m, float* b, size_t bstep, int n);
+CV_EXPORTS bool Cholesky(double* A, size_t astep, int m, double* b, size_t bstep, int n);
+
+CV_EXPORTS void exp(const float* src, float* dst, int n);
+CV_EXPORTS void exp(const double* src, double* dst, int n);
+CV_EXPORTS void log(const float* src, float* dst, int n);
+CV_EXPORTS void log(const double* src, double* dst, int n);
+
+CV_EXPORTS void fastAtan2(const float* y, const float* x, float* dst, int n, bool angleInDegrees);
+CV_EXPORTS void magnitude(const float* x, const float* y, float* dst, int n);
+CV_EXPORTS void magnitude(const double* x, const double* y, double* dst, int n);
+CV_EXPORTS void sqrt(const float* src, float* dst, int len);
+CV_EXPORTS void sqrt(const double* src, double* dst, int len);
+CV_EXPORTS void invSqrt(const float* src, float* dst, int len);
+CV_EXPORTS void invSqrt(const double* src, double* dst, int len);
+
+//! @endcond
+
+}} //cv::hal
+
+#endif //OPENCV_HAL_HPP

model_zoo/official/lite/app/src/main/cpp/include/opencv2/core/hal/interface.h

@@ -0,0 +1,190 @@
+#ifndef OPENCV_CORE_HAL_INTERFACE_H
+#define OPENCV_CORE_HAL_INTERFACE_H
+
+//! @addtogroup core_hal_interface
+//! @{
+
+//! @name Return codes
+//! @{
+#define CV_HAL_ERROR_OK 0
+#define CV_HAL_ERROR_NOT_IMPLEMENTED 1
+#define CV_HAL_ERROR_UNKNOWN -1
+//! @}
+
+#ifdef __cplusplus
+#include <cstddef>
+#else
+#include <stddef.h>
+#include <stdbool.h>
+#endif
+
+//! @name Data types
+//! primitive types
+//! - schar  - signed 1 byte integer
+//! - uchar  - unsigned 1 byte integer
+//! - short  - signed 2 byte integer
+//! - ushort - unsigned 2 byte integer
+//! - int    - signed 4 byte integer
+//! - uint   - unsigned 4 byte integer
+//! - int64  - signed 8 byte integer
+//! - uint64 - unsigned 8 byte integer
+//! @{
+#if !defined _MSC_VER && !defined __BORLANDC__
+#  if defined __cplusplus && __cplusplus >= 201103L && !defined __APPLE__
+#    include <cstdint>
+#    ifdef __NEWLIB__
+        typedef unsigned int uint;
+#    else
+        typedef std::uint32_t uint;
+#    endif
+#  else
+#    include <stdint.h>
+     typedef uint32_t uint;
+#  endif
+#else
+   typedef unsigned uint;
+#endif
+
+typedef signed char schar;
+
+#ifndef __IPL_H__
+   typedef unsigned char uchar;
+   typedef unsigned short ushort;
+#endif
+
+#if defined _MSC_VER || defined __BORLANDC__
+   typedef __int64 int64;
+   typedef unsigned __int64 uint64;
+#  define CV_BIG_INT(n)   n##I64
+#  define CV_BIG_UINT(n)  n##UI64
+#else
+   typedef int64_t int64;
+   typedef uint64_t uint64;
+#  define CV_BIG_INT(n)   n##LL
+#  define CV_BIG_UINT(n)  n##ULL
+#endif
+
+#define CV_USRTYPE1 (void)"CV_USRTYPE1 support has been dropped in OpenCV 4.0"
+
+#define CV_CN_MAX     512
+#define CV_CN_SHIFT   3
+#define CV_DEPTH_MAX  (1 << CV_CN_SHIFT)
+
+#define CV_8U   0
+#define CV_8S   1
+#define CV_16U  2
+#define CV_16S  3
+#define CV_32S  4
+#define CV_32F  5
+#define CV_64F  6
+#define CV_16F  7
+
+#define CV_MAT_DEPTH_MASK       (CV_DEPTH_MAX - 1)
+#define CV_MAT_DEPTH(flags)     ((flags) & CV_MAT_DEPTH_MASK)
+
+#define CV_MAKETYPE(depth,cn) (CV_MAT_DEPTH(depth) + (((cn)-1) << CV_CN_SHIFT))
+#define CV_MAKE_TYPE CV_MAKETYPE
+
+#define CV_8UC1 CV_MAKETYPE(CV_8U,1)
+#define CV_8UC2 CV_MAKETYPE(CV_8U,2)
+#define CV_8UC3 CV_MAKETYPE(CV_8U,3)
+#define CV_8UC4 CV_MAKETYPE(CV_8U,4)
+#define CV_8UC(n) CV_MAKETYPE(CV_8U,(n))
+
+#define CV_8SC1 CV_MAKETYPE(CV_8S,1)
+#define CV_8SC2 CV_MAKETYPE(CV_8S,2)
+#define CV_8SC3 CV_MAKETYPE(CV_8S,3)
+#define CV_8SC4 CV_MAKETYPE(CV_8S,4)
+#define CV_8SC(n) CV_MAKETYPE(CV_8S,(n))
+
+#define CV_16UC1 CV_MAKETYPE(CV_16U,1)
+#define CV_16UC2 CV_MAKETYPE(CV_16U,2)
+#define CV_16UC3 CV_MAKETYPE(CV_16U,3)
+#define CV_16UC4 CV_MAKETYPE(CV_16U,4)
+#define CV_16UC(n) CV_MAKETYPE(CV_16U,(n))
+
+#define CV_16SC1 CV_MAKETYPE(CV_16S,1)
+#define CV_16SC2 CV_MAKETYPE(CV_16S,2)
+#define CV_16SC3 CV_MAKETYPE(CV_16S,3)
+#define CV_16SC4 CV_MAKETYPE(CV_16S,4)
+#define CV_16SC(n) CV_MAKETYPE(CV_16S,(n))
+
+#define CV_32SC1 CV_MAKETYPE(CV_32S,1)
+#define CV_32SC2 CV_MAKETYPE(CV_32S,2)
+#define CV_32SC3 CV_MAKETYPE(CV_32S,3)
+#define CV_32SC4 CV_MAKETYPE(CV_32S,4)
+#define CV_32SC(n) CV_MAKETYPE(CV_32S,(n))
+
+#define CV_32FC1 CV_MAKETYPE(CV_32F,1)
+#define CV_32FC2 CV_MAKETYPE(CV_32F,2)
+#define CV_32FC3 CV_MAKETYPE(CV_32F,3)
+#define CV_32FC4 CV_MAKETYPE(CV_32F,4)
+#define CV_32FC(n) CV_MAKETYPE(CV_32F,(n))
+
+#define CV_64FC1 CV_MAKETYPE(CV_64F,1)
+#define CV_64FC2 CV_MAKETYPE(CV_64F,2)
+#define CV_64FC3 CV_MAKETYPE(CV_64F,3)
+#define CV_64FC4 CV_MAKETYPE(CV_64F,4)
+#define CV_64FC(n) CV_MAKETYPE(CV_64F,(n))
+
+#define CV_16FC1 CV_MAKETYPE(CV_16F,1)
+#define CV_16FC2 CV_MAKETYPE(CV_16F,2)
+#define CV_16FC3 CV_MAKETYPE(CV_16F,3)
+#define CV_16FC4 CV_MAKETYPE(CV_16F,4)
+#define CV_16FC(n) CV_MAKETYPE(CV_16F,(n))
+//! @}
+
+//! @name Comparison operation
+//! @sa cv::CmpTypes
+//! @{
+#define CV_HAL_CMP_EQ 0
+#define CV_HAL_CMP_GT 1
+#define CV_HAL_CMP_GE 2
+#define CV_HAL_CMP_LT 3
+#define CV_HAL_CMP_LE 4
+#define CV_HAL_CMP_NE 5
+//! @}
+
+//! @name Border processing modes
+//! @sa cv::BorderTypes
+//! @{
+#define CV_HAL_BORDER_CONSTANT 0
+#define CV_HAL_BORDER_REPLICATE 1
+#define CV_HAL_BORDER_REFLECT 2
+#define CV_HAL_BORDER_WRAP 3
+#define CV_HAL_BORDER_REFLECT_101 4
+#define CV_HAL_BORDER_TRANSPARENT 5
+#define CV_HAL_BORDER_ISOLATED 16
+//! @}
+
+//! @name DFT flags
+//! @{
+#define CV_HAL_DFT_INVERSE        1
+#define CV_HAL_DFT_SCALE          2
+#define CV_HAL_DFT_ROWS           4
+#define CV_HAL_DFT_COMPLEX_OUTPUT 16
+#define CV_HAL_DFT_REAL_OUTPUT    32
+#define CV_HAL_DFT_TWO_STAGE      64
+#define CV_HAL_DFT_STAGE_COLS    128
+#define CV_HAL_DFT_IS_CONTINUOUS 512
+#define CV_HAL_DFT_IS_INPLACE 1024
+//! @}
+
+//! @name SVD flags
+//! @{
+#define CV_HAL_SVD_NO_UV    1
+#define CV_HAL_SVD_SHORT_UV 2
+#define CV_HAL_SVD_MODIFY_A 4
+#define CV_HAL_SVD_FULL_UV  8
+//! @}
+
+//! @name Gemm flags
+//! @{
+#define CV_HAL_GEMM_1_T 1
+#define CV_HAL_GEMM_2_T 2
+#define CV_HAL_GEMM_3_T 4
+//! @}
+
+//! @}
+
+#endif

model_zoo/official/lite/app/src/main/cpp/include/opencv2/core/hal/intrin.hpp

@@ -0,0 +1,520 @@
+/*M///////////////////////////////////////////////////////////////////////////////////////
+//
+//  IMPORTANT: READ BEFORE DOWNLOADING, COPYING, INSTALLING OR USING.
+//
+//  By downloading, copying, installing or using the software you agree to this license.
+//  If you do not agree to this license, do not download, install,
+//  copy or use the software.
+//
+//
+//                          License Agreement
+//                For Open Source Computer Vision Library
+//
+// Copyright (C) 2000-2008, Intel Corporation, all rights reserved.
+// Copyright (C) 2009, Willow Garage Inc., all rights reserved.
+// Copyright (C) 2013, OpenCV Foundation, all rights reserved.
+// Copyright (C) 2015, Itseez Inc., all rights reserved.
+// Third party copyrights are property of their respective owners.
+//
+// Redistribution and use in source and binary forms, with or without modification,
+// are permitted provided that the following conditions are met:
+//
+//   * Redistribution's of source code must retain the above copyright notice,
+//     this list of conditions and the following disclaimer.
+//
+//   * Redistribution's in binary form must reproduce the above copyright notice,
+//     this list of conditions and the following disclaimer in the documentation
+//     and/or other materials provided with the distribution.
+//
+//   * The name of the copyright holders may not be used to endorse or promote products
+//     derived from this software without specific prior written permission.
+//
+// This software is provided by the copyright holders and contributors "as is" and
+// any express or implied warranties, including, but not limited to, the implied
+// warranties of merchantability and fitness for a particular purpose are disclaimed.
+// In no event shall the Intel Corporation or contributors be liable for any direct,
+// indirect, incidental, special, exemplary, or consequential damages
+// (including, but not limited to, procurement of substitute goods or services;
+// loss of use, data, or profits; or business interruption) however caused
+// and on any theory of liability, whether in contract, strict liability,
+// or tort (including negligence or otherwise) arising in any way out of
+// the use of this software, even if advised of the possibility of such damage.
+//
+//M*/
+
+#ifndef OPENCV_HAL_INTRIN_HPP
+#define OPENCV_HAL_INTRIN_HPP
+
+#include <cmath>
+#include <float.h>
+#include <stdlib.h>
+#include "opencv2/core/cvdef.h"
+
+#define OPENCV_HAL_ADD(a, b) ((a) + (b))
+#define OPENCV_HAL_AND(a, b) ((a) & (b))
+#define OPENCV_HAL_NOP(a) (a)
+#define OPENCV_HAL_1ST(a, b) (a)
+
+namespace {
+inline unsigned int trailingZeros32(unsigned int value) {
+#if defined(_MSC_VER)
+#if (_MSC_VER < 1700) || defined(_M_ARM) || defined(_M_ARM64)
+    unsigned long index = 0;
+    _BitScanForward(&index, value);
+    return (unsigned int)index;
+#elif defined(__clang__)
+    // clang-cl doesn't export _tzcnt_u32 for non BMI systems
+    return value ? __builtin_ctz(value) : 32;
+#else
+    return _tzcnt_u32(value);
+#endif
+#elif defined(__GNUC__) || defined(__GNUG__)
+    return __builtin_ctz(value);
+#elif defined(__ICC) || defined(__INTEL_COMPILER)
+    return _bit_scan_forward(value);
+#elif defined(__clang__)
+    return llvm.cttz.i32(value, true);
+#else
+    static const int MultiplyDeBruijnBitPosition[32] = {
+        0, 1, 28, 2, 29, 14, 24, 3, 30, 22, 20, 15, 25, 17, 4, 8,
+        31, 27, 13, 23, 21, 19, 16, 7, 26, 12, 18, 6, 11, 5, 10, 9 };
+    return MultiplyDeBruijnBitPosition[((uint32_t)((value & -value) * 0x077CB531U)) >> 27];
+#endif
+}
+}
+
+// unlike HAL API, which is in cv::hal,
+// we put intrinsics into cv namespace to make its
+// access from within opencv code more accessible
+namespace cv {
+
+namespace hal {
+
+enum StoreMode
+{
+    STORE_UNALIGNED = 0,
+    STORE_ALIGNED = 1,
+    STORE_ALIGNED_NOCACHE = 2
+};
+
+}
+
+// TODO FIXIT: Don't use "God" traits. Split on separate cases.
+template<typename _Tp> struct V_TypeTraits
+{
+};
+
+#define CV_INTRIN_DEF_TYPE_TRAITS(type, int_type_, uint_type_, abs_type_, w_type_, q_type_, sum_type_, nlanes128_) \
+    template<> struct V_TypeTraits<type> \
+    { \
+        typedef type value_type; \
+        typedef int_type_ int_type; \
+        typedef abs_type_ abs_type; \
+        typedef uint_type_ uint_type; \
+        typedef w_type_ w_type; \
+        typedef q_type_ q_type; \
+        typedef sum_type_ sum_type; \
+        enum { nlanes128 = nlanes128_ }; \
+    \
+        static inline int_type reinterpret_int(type x) \
+        { \
+            union { type l; int_type i; } v; \
+            v.l = x; \
+            return v.i; \
+        } \
+    \
+        static inline type reinterpret_from_int(int_type x) \
+        { \
+            union { type l; int_type i; } v; \
+            v.i = x; \
+            return v.l; \
+        } \
+    }
+
+#define CV_INTRIN_DEF_TYPE_TRAITS_NO_Q_TYPE(type, int_type_, uint_type_, abs_type_, w_type_, sum_type_, nlanes128_) \
+    template<> struct V_TypeTraits<type> \
+    { \
+        typedef type value_type; \
+        typedef int_type_ int_type; \
+        typedef abs_type_ abs_type; \
+        typedef uint_type_ uint_type; \
+        typedef w_type_ w_type; \
+        typedef sum_type_ sum_type; \
+        enum { nlanes128 = nlanes128_ }; \
+    \
+        static inline int_type reinterpret_int(type x) \
+        { \
+            union { type l; int_type i; } v; \
+            v.l = x; \
+            return v.i; \
+        } \
+    \
+        static inline type reinterpret_from_int(int_type x) \
+        { \
+            union { type l; int_type i; } v; \
+            v.i = x; \
+            return v.l; \
+        } \
+    }
+
+CV_INTRIN_DEF_TYPE_TRAITS(uchar, schar, uchar, uchar, ushort, unsigned, unsigned, 16);
+CV_INTRIN_DEF_TYPE_TRAITS(schar, schar, uchar, uchar, short, int, int, 16);
+CV_INTRIN_DEF_TYPE_TRAITS(ushort, short, ushort, ushort, unsigned, uint64, unsigned, 8);
+CV_INTRIN_DEF_TYPE_TRAITS(short, short, ushort, ushort, int, int64, int, 8);
+CV_INTRIN_DEF_TYPE_TRAITS_NO_Q_TYPE(unsigned, int, unsigned, unsigned, uint64, unsigned, 4);
+CV_INTRIN_DEF_TYPE_TRAITS_NO_Q_TYPE(int, int, unsigned, unsigned, int64, int, 4);
+CV_INTRIN_DEF_TYPE_TRAITS_NO_Q_TYPE(float, int, unsigned, float, double, float, 4);
+CV_INTRIN_DEF_TYPE_TRAITS_NO_Q_TYPE(uint64, int64, uint64, uint64, void, uint64, 2);
+CV_INTRIN_DEF_TYPE_TRAITS_NO_Q_TYPE(int64, int64, uint64, uint64, void, int64, 2);
+CV_INTRIN_DEF_TYPE_TRAITS_NO_Q_TYPE(double, int64, uint64, double, void, double, 2);
+
+#ifndef CV_DOXYGEN
+
+#ifndef CV_CPU_OPTIMIZATION_HAL_NAMESPACE
+#ifdef CV_FORCE_SIMD128_CPP
+    #define CV_CPU_OPTIMIZATION_HAL_NAMESPACE hal_EMULATOR_CPP
+    #define CV_CPU_OPTIMIZATION_HAL_NAMESPACE_BEGIN namespace hal_EMULATOR_CPP {
+    #define CV_CPU_OPTIMIZATION_HAL_NAMESPACE_END }
+#elif defined(CV_CPU_DISPATCH_MODE)
+    #define CV_CPU_OPTIMIZATION_HAL_NAMESPACE __CV_CAT(hal_, CV_CPU_DISPATCH_MODE)
+    #define CV_CPU_OPTIMIZATION_HAL_NAMESPACE_BEGIN namespace __CV_CAT(hal_, CV_CPU_DISPATCH_MODE) {
+    #define CV_CPU_OPTIMIZATION_HAL_NAMESPACE_END }
+#else
+    #define CV_CPU_OPTIMIZATION_HAL_NAMESPACE hal_baseline
+    #define CV_CPU_OPTIMIZATION_HAL_NAMESPACE_BEGIN namespace hal_baseline {
+    #define CV_CPU_OPTIMIZATION_HAL_NAMESPACE_END }
+#endif
+#endif // CV_CPU_OPTIMIZATION_HAL_NAMESPACE
+
+CV_CPU_OPTIMIZATION_HAL_NAMESPACE_BEGIN
+CV_CPU_OPTIMIZATION_HAL_NAMESPACE_END
+using namespace CV_CPU_OPTIMIZATION_HAL_NAMESPACE;
+#endif
+}
+
+#ifdef CV_DOXYGEN
+#   undef CV_AVX2
+#   undef CV_SSE2
+#   undef CV_NEON
+#   undef CV_VSX
+#   undef CV_FP16
+#   undef CV_MSA
+#endif
+
+#if (CV_SSE2 || CV_NEON || CV_VSX || CV_MSA || CV_WASM_SIMD) && !defined(CV_FORCE_SIMD128_CPP)
+#define CV__SIMD_FORWARD 128
+#include "opencv2/core/hal/intrin_forward.hpp"
+#endif
+
+#if CV_SSE2 && !defined(CV_FORCE_SIMD128_CPP)
+
+#include "opencv2/core/hal/intrin_sse_em.hpp"
+#include "opencv2/core/hal/intrin_sse.hpp"
+
+#elif CV_NEON && !defined(CV_FORCE_SIMD128_CPP)
+
+#include "opencv2/core/hal/intrin_neon.hpp"
+
+#elif CV_VSX && !defined(CV_FORCE_SIMD128_CPP)
+
+#include "opencv2/core/hal/intrin_vsx.hpp"
+
+#elif CV_MSA && !defined(CV_FORCE_SIMD128_CPP)
+
+#include "opencv2/core/hal/intrin_msa.hpp"
+
+#elif CV_WASM_SIMD && !defined(CV_FORCE_SIMD128_CPP)
+#include "opencv2/core/hal/intrin_wasm.hpp"
+
+#else
+
+#include "opencv2/core/hal/intrin_cpp.hpp"
+
+#endif
+
+// AVX2 can be used together with SSE2, so
+// we define those two sets of intrinsics at once.
+// Most of the intrinsics do not conflict (the proper overloaded variant is
+// resolved by the argument types, e.g. v_float32x4 ~ SSE2, v_float32x8 ~ AVX2),
+// but some of AVX2 intrinsics get v256_ prefix instead of v_, e.g. v256_load() vs v_load().
+// Correspondingly, the wide intrinsics (which are mapped to the "widest"
+// available instruction set) will get vx_ prefix
+// (and will be mapped to v256_ counterparts) (e.g. vx_load() => v256_load())
+#if CV_AVX2
+
+#define CV__SIMD_FORWARD 256
+#include "opencv2/core/hal/intrin_forward.hpp"
+#include "opencv2/core/hal/intrin_avx.hpp"
+
+#endif
+
+// AVX512 can be used together with SSE2 and AVX2, so
+// we define those sets of intrinsics at once.
+// For some of AVX512 intrinsics get v512_ prefix instead of v_, e.g. v512_load() vs v_load().
+// Wide intrinsics will be mapped to v512_ counterparts in this case(e.g. vx_load() => v512_load())
+#if CV_AVX512_SKX
+
+#define CV__SIMD_FORWARD 512
+#include "opencv2/core/hal/intrin_forward.hpp"
+#include "opencv2/core/hal/intrin_avx512.hpp"
+
+#endif
+
+//! @cond IGNORED
+
+namespace cv {
+
+#ifndef CV_DOXYGEN
+CV_CPU_OPTIMIZATION_HAL_NAMESPACE_BEGIN
+#endif
+
+#ifndef CV_SIMD128
+#define CV_SIMD128 0
+#endif
+
+#ifndef CV_SIMD128_CPP
+#define CV_SIMD128_CPP 0
+#endif
+
+#ifndef CV_SIMD128_64F
+#define CV_SIMD128_64F 0
+#endif
+
+#ifndef CV_SIMD256
+#define CV_SIMD256 0
+#endif
+
+#ifndef CV_SIMD256_64F
+#define CV_SIMD256_64F 0
+#endif
+
+#ifndef CV_SIMD512
+#define CV_SIMD512 0
+#endif
+
+#ifndef CV_SIMD512_64F
+#define CV_SIMD512_64F 0
+#endif
+
+#ifndef CV_SIMD128_FP16
+#define CV_SIMD128_FP16 0
+#endif
+
+#ifndef CV_SIMD256_FP16
+#define CV_SIMD256_FP16 0
+#endif
+
+#ifndef CV_SIMD512_FP16
+#define CV_SIMD512_FP16 0
+#endif
+
+//==================================================================================================
+
+#define CV_INTRIN_DEFINE_WIDE_INTRIN(typ, vtyp, short_typ, prefix, loadsfx) \
+    inline vtyp vx_setall_##short_typ(typ v) { return prefix##_setall_##short_typ(v); } \
+    inline vtyp vx_setzero_##short_typ() { return prefix##_setzero_##short_typ(); } \
+    inline vtyp vx_##loadsfx(const typ* ptr) { return prefix##_##loadsfx(ptr); } \
+    inline vtyp vx_##loadsfx##_aligned(const typ* ptr) { return prefix##_##loadsfx##_aligned(ptr); } \
+    inline vtyp vx_##loadsfx##_low(const typ* ptr) { return prefix##_##loadsfx##_low(ptr); } \
+    inline vtyp vx_##loadsfx##_halves(const typ* ptr0, const typ* ptr1) { return prefix##_##loadsfx##_halves(ptr0, ptr1); } \
+    inline void vx_store(typ* ptr, const vtyp& v) { return v_store(ptr, v); } \
+    inline void vx_store_aligned(typ* ptr, const vtyp& v) { return v_store_aligned(ptr, v); } \
+    inline vtyp vx_lut(const typ* ptr, const int* idx) { return prefix##_lut(ptr, idx); } \
+    inline vtyp vx_lut_pairs(const typ* ptr, const int* idx) { return prefix##_lut_pairs(ptr, idx); }
+
+#define CV_INTRIN_DEFINE_WIDE_LUT_QUAD(typ, vtyp, prefix) \
+    inline vtyp vx_lut_quads(const typ* ptr, const int* idx) { return prefix##_lut_quads(ptr, idx); }
+
+#define CV_INTRIN_DEFINE_WIDE_LOAD_EXPAND(typ, wtyp, prefix) \
+    inline wtyp vx_load_expand(const typ* ptr) { return prefix##_load_expand(ptr); }
+
+#define CV_INTRIN_DEFINE_WIDE_LOAD_EXPAND_Q(typ, qtyp, prefix) \
+    inline qtyp vx_load_expand_q(const typ* ptr) { return prefix##_load_expand_q(ptr); }
+
+#define CV_INTRIN_DEFINE_WIDE_INTRIN_WITH_EXPAND(typ, vtyp, short_typ, wtyp, qtyp, prefix, loadsfx) \
+    CV_INTRIN_DEFINE_WIDE_INTRIN(typ, vtyp, short_typ, prefix, loadsfx) \
+    CV_INTRIN_DEFINE_WIDE_LUT_QUAD(typ, vtyp, prefix) \
+    CV_INTRIN_DEFINE_WIDE_LOAD_EXPAND(typ, wtyp, prefix) \
+    CV_INTRIN_DEFINE_WIDE_LOAD_EXPAND_Q(typ, qtyp, prefix)
+
+#define CV_INTRIN_DEFINE_WIDE_INTRIN_ALL_TYPES(prefix) \
+    CV_INTRIN_DEFINE_WIDE_INTRIN_WITH_EXPAND(uchar, v_uint8, u8, v_uint16, v_uint32, prefix, load) \
+    CV_INTRIN_DEFINE_WIDE_INTRIN_WITH_EXPAND(schar, v_int8, s8, v_int16, v_int32, prefix, load) \
+    CV_INTRIN_DEFINE_WIDE_INTRIN(ushort, v_uint16, u16, prefix, load) \
+    CV_INTRIN_DEFINE_WIDE_LUT_QUAD(ushort, v_uint16, prefix) \
+    CV_INTRIN_DEFINE_WIDE_LOAD_EXPAND(ushort, v_uint32, prefix) \
+    CV_INTRIN_DEFINE_WIDE_INTRIN(short, v_int16, s16, prefix, load) \
+    CV_INTRIN_DEFINE_WIDE_LUT_QUAD(short, v_int16, prefix) \
+    CV_INTRIN_DEFINE_WIDE_LOAD_EXPAND(short, v_int32, prefix) \
+    CV_INTRIN_DEFINE_WIDE_INTRIN(int, v_int32, s32, prefix, load) \
+    CV_INTRIN_DEFINE_WIDE_LUT_QUAD(int, v_int32, prefix) \
+    CV_INTRIN_DEFINE_WIDE_LOAD_EXPAND(int, v_int64, prefix) \
+    CV_INTRIN_DEFINE_WIDE_INTRIN(unsigned, v_uint32, u32, prefix, load) \
+    CV_INTRIN_DEFINE_WIDE_LUT_QUAD(unsigned, v_uint32, prefix) \
+    CV_INTRIN_DEFINE_WIDE_LOAD_EXPAND(unsigned, v_uint64, prefix) \
+    CV_INTRIN_DEFINE_WIDE_INTRIN(float, v_float32, f32, prefix, load) \
+    CV_INTRIN_DEFINE_WIDE_LUT_QUAD(float, v_float32, prefix) \
+    CV_INTRIN_DEFINE_WIDE_INTRIN(int64, v_int64, s64, prefix, load) \
+    CV_INTRIN_DEFINE_WIDE_INTRIN(uint64, v_uint64, u64, prefix, load) \
+    CV_INTRIN_DEFINE_WIDE_LOAD_EXPAND(float16_t, v_float32, prefix)
+
+template<typename _Tp> struct V_RegTraits
+{
+};
+
+#define CV_DEF_REG_TRAITS(prefix, _reg, lane_type, suffix, _u_reg, _w_reg, _q_reg, _int_reg, _round_reg) \
+    template<> struct V_RegTraits<_reg> \
+    { \
+        typedef _reg reg; \
+        typedef _u_reg u_reg; \
+        typedef _w_reg w_reg; \
+        typedef _q_reg q_reg; \
+        typedef _int_reg int_reg; \
+        typedef _round_reg round_reg; \
+    }
+
+#if CV_SIMD128 || CV_SIMD128_CPP
+    CV_DEF_REG_TRAITS(v, v_uint8x16, uchar, u8, v_uint8x16, v_uint16x8, v_uint32x4, v_int8x16, void);
+    CV_DEF_REG_TRAITS(v, v_int8x16, schar, s8, v_uint8x16, v_int16x8, v_int32x4, v_int8x16, void);
+    CV_DEF_REG_TRAITS(v, v_uint16x8, ushort, u16, v_uint16x8, v_uint32x4, v_uint64x2, v_int16x8, void);
+    CV_DEF_REG_TRAITS(v, v_int16x8, short, s16, v_uint16x8, v_int32x4, v_int64x2, v_int16x8, void);
+    CV_DEF_REG_TRAITS(v, v_uint32x4, unsigned, u32, v_uint32x4, v_uint64x2, void, v_int32x4, void);
+    CV_DEF_REG_TRAITS(v, v_int32x4, int, s32, v_uint32x4, v_int64x2, void, v_int32x4, void);
+#if CV_SIMD128_64F || CV_SIMD128_CPP
+    CV_DEF_REG_TRAITS(v, v_float32x4, float, f32, v_float32x4, v_float64x2, void, v_int32x4, v_int32x4);
+#else
+    CV_DEF_REG_TRAITS(v, v_float32x4, float, f32, v_float32x4, void, void, v_int32x4, v_int32x4);
+#endif
+    CV_DEF_REG_TRAITS(v, v_uint64x2, uint64, u64, v_uint64x2, void, void, v_int64x2, void);
+    CV_DEF_REG_TRAITS(v, v_int64x2, int64, s64, v_uint64x2, void, void, v_int64x2, void);
+#if CV_SIMD128_64F
+    CV_DEF_REG_TRAITS(v, v_float64x2, double, f64, v_float64x2, void, void, v_int64x2, v_int32x4);
+#endif
+#endif
+
+#if CV_SIMD256
+    CV_DEF_REG_TRAITS(v256, v_uint8x32, uchar, u8, v_uint8x32, v_uint16x16, v_uint32x8, v_int8x32, void);
+    CV_DEF_REG_TRAITS(v256, v_int8x32, schar, s8, v_uint8x32, v_int16x16, v_int32x8, v_int8x32, void);
+    CV_DEF_REG_TRAITS(v256, v_uint16x16, ushort, u16, v_uint16x16, v_uint32x8, v_uint64x4, v_int16x16, void);
+    CV_DEF_REG_TRAITS(v256, v_int16x16, short, s16, v_uint16x16, v_int32x8, v_int64x4, v_int16x16, void);
+    CV_DEF_REG_TRAITS(v256, v_uint32x8, unsigned, u32, v_uint32x8, v_uint64x4, void, v_int32x8, void);
+    CV_DEF_REG_TRAITS(v256, v_int32x8, int, s32, v_uint32x8, v_int64x4, void, v_int32x8, void);
+    CV_DEF_REG_TRAITS(v256, v_float32x8, float, f32, v_float32x8, v_float64x4, void, v_int32x8, v_int32x8);
+    CV_DEF_REG_TRAITS(v256, v_uint64x4, uint64, u64, v_uint64x4, void, void, v_int64x4, void);
+    CV_DEF_REG_TRAITS(v256, v_int64x4, int64, s64, v_uint64x4, void, void, v_int64x4, void);
+    CV_DEF_REG_TRAITS(v256, v_float64x4, double, f64, v_float64x4, void, void, v_int64x4, v_int32x8);
+#endif
+
+#if CV_SIMD512
+    CV_DEF_REG_TRAITS(v512, v_uint8x64, uchar, u8, v_uint8x64, v_uint16x32, v_uint32x16, v_int8x64, void);
+    CV_DEF_REG_TRAITS(v512, v_int8x64, schar, s8, v_uint8x64, v_int16x32, v_int32x16, v_int8x64, void);
+    CV_DEF_REG_TRAITS(v512, v_uint16x32, ushort, u16, v_uint16x32, v_uint32x16, v_uint64x8, v_int16x32, void);
+    CV_DEF_REG_TRAITS(v512, v_int16x32, short, s16, v_uint16x32, v_int32x16, v_int64x8, v_int16x32, void);
+    CV_DEF_REG_TRAITS(v512, v_uint32x16, unsigned, u32, v_uint32x16, v_uint64x8, void, v_int32x16, void);
+    CV_DEF_REG_TRAITS(v512, v_int32x16, int, s32, v_uint32x16, v_int64x8, void, v_int32x16, void);
+    CV_DEF_REG_TRAITS(v512, v_float32x16, float, f32, v_float32x16, v_float64x8, void, v_int32x16, v_int32x16);
+    CV_DEF_REG_TRAITS(v512, v_uint64x8, uint64, u64, v_uint64x8, void, void, v_int64x8, void);
+    CV_DEF_REG_TRAITS(v512, v_int64x8, int64, s64, v_uint64x8, void, void, v_int64x8, void);
+    CV_DEF_REG_TRAITS(v512, v_float64x8, double, f64, v_float64x8, void, void, v_int64x8, v_int32x16);
+#endif
+
+#if CV_SIMD512 && (!defined(CV__SIMD_FORCE_WIDTH) || CV__SIMD_FORCE_WIDTH == 512)
+#define CV__SIMD_NAMESPACE simd512
+namespace CV__SIMD_NAMESPACE {
+    #define CV_SIMD 1
+    #define CV_SIMD_64F CV_SIMD512_64F
+    #define CV_SIMD_FP16 CV_SIMD512_FP16
+    #define CV_SIMD_WIDTH 64
+    typedef v_uint8x64    v_uint8;
+    typedef v_int8x64     v_int8;
+    typedef v_uint16x32   v_uint16;
+    typedef v_int16x32    v_int16;
+    typedef v_uint32x16   v_uint32;
+    typedef v_int32x16    v_int32;
+    typedef v_uint64x8    v_uint64;
+    typedef v_int64x8     v_int64;
+    typedef v_float32x16  v_float32;
+    CV_INTRIN_DEFINE_WIDE_INTRIN_ALL_TYPES(v512)
+#if CV_SIMD512_64F
+    typedef v_float64x8   v_float64;
+    CV_INTRIN_DEFINE_WIDE_INTRIN(double, v_float64, f64, v512, load)
+#endif
+        inline void vx_cleanup() { v512_cleanup(); }
+} // namespace
+using namespace CV__SIMD_NAMESPACE;
+#elif CV_SIMD256 && (!defined(CV__SIMD_FORCE_WIDTH) || CV__SIMD_FORCE_WIDTH == 256)
+#define CV__SIMD_NAMESPACE simd256
+namespace CV__SIMD_NAMESPACE {
+    #define CV_SIMD 1
+    #define CV_SIMD_64F CV_SIMD256_64F
+    #define CV_SIMD_FP16 CV_SIMD256_FP16
+    #define CV_SIMD_WIDTH 32
+    typedef v_uint8x32   v_uint8;
+    typedef v_int8x32    v_int8;
+    typedef v_uint16x16  v_uint16;
+    typedef v_int16x16   v_int16;
+    typedef v_uint32x8   v_uint32;
+    typedef v_int32x8    v_int32;
+    typedef v_uint64x4   v_uint64;
+    typedef v_int64x4    v_int64;
+    typedef v_float32x8  v_float32;
+    CV_INTRIN_DEFINE_WIDE_INTRIN_ALL_TYPES(v256)
+    #if CV_SIMD256_64F
+    typedef v_float64x4  v_float64;
+    CV_INTRIN_DEFINE_WIDE_INTRIN(double, v_float64, f64, v256, load)
+    #endif
+    inline void vx_cleanup() { v256_cleanup(); }
+} // namespace
+using namespace CV__SIMD_NAMESPACE;
+#elif (CV_SIMD128 || CV_SIMD128_CPP) && (!defined(CV__SIMD_FORCE_WIDTH) || CV__SIMD_FORCE_WIDTH == 128)
+#if defined CV_SIMD128_CPP
+#define CV__SIMD_NAMESPACE simd128_cpp
+#else
+#define CV__SIMD_NAMESPACE simd128
+#endif
+namespace CV__SIMD_NAMESPACE {
+    #define CV_SIMD CV_SIMD128
+    #define CV_SIMD_64F CV_SIMD128_64F
+    #define CV_SIMD_WIDTH 16
+    typedef v_uint8x16  v_uint8;
+    typedef v_int8x16   v_int8;
+    typedef v_uint16x8  v_uint16;
+    typedef v_int16x8   v_int16;
+    typedef v_uint32x4  v_uint32;
+    typedef v_int32x4   v_int32;
+    typedef v_uint64x2  v_uint64;
+    typedef v_int64x2   v_int64;
+    typedef v_float32x4 v_float32;
+    CV_INTRIN_DEFINE_WIDE_INTRIN_ALL_TYPES(v)
+    #if CV_SIMD128_64F
+    typedef v_float64x2 v_float64;
+    CV_INTRIN_DEFINE_WIDE_INTRIN(double, v_float64, f64, v, load)
+    #endif
+    inline void vx_cleanup() { v_cleanup(); }
+} // namespace
+using namespace CV__SIMD_NAMESPACE;
+#endif
+
+#ifndef CV_SIMD_64F
+#define CV_SIMD_64F 0
+#endif
+
+#ifndef CV_SIMD_FP16
+#define CV_SIMD_FP16 0  //!< Defined to 1 on native support of operations with float16x8_t / float16x16_t (SIMD256) types
+#endif
+
+#ifndef CV_SIMD
+#define CV_SIMD 0
+#endif
+
+#include "simd_utils.impl.hpp"
+
+#ifndef CV_DOXYGEN
+CV_CPU_OPTIMIZATION_HAL_NAMESPACE_END
+#endif
+
+} // cv::
+
+//! @endcond
+
+#endif