12 kB
- 如何加入技术交流QQ群？
- 如何看作者b站直播？
- 编译
-
- 怎样下载源码？
- 怎么交叉编译？
- The submodules were not downloaded! Please update submodules with "git submodule update --init" and try again
- Could NOT find Protobuf (missing: Protobuf_INCLUDE_DIR)
- Could NOT find CUDA (missing: CUDA_TOOLKIT_ROOT_DIR CUDA_INCLUDE_DIRS CUDA_CUDART_LIBRARY)
- Could not find a package configuration file provided by "OpenCV" with any of the following names: OpenCVConfig.cmake opencv-config.cmake
- Could not find a package configuration file provided by "ncnn" with any of the following names: ncnnConfig.cmake ncnn-config.cmake
- 找不到 Vulkan
- undefined reference to __kmpc_for_static_init_4 __kmpc_for_static_fini __kmpc_fork_call ...
- undefined reference to vkEnumerateInstanceExtensionProperties vkGetInstanceProcAddr vkQueueSubmit ...
- undefined reference to glslang::InitializeProcess() glslang::TShader::TShader(EShLanguage) ...
- undefined reference to AAssetManager_fromJava AAssetManager_open AAsset_seek ...
- undefined reference to typeinfo for ncnn::Layer
- undefined reference to __cpu_model
- unrecognized command line option "-mavx2"
- 为啥自己编译的ncnn android库特别大？
- rtti/exceptions冲突
- 怎样添加ncnn库到项目中？cmake方式怎么用？
- 转模型问题
-
- caffe
- mxnet
- darknet
- pytorch - onnx
- tensorflow 1.x/2.x - keras
- tensorflow 2.x - mlir
- Shape not supported yet! Gather not supported yet! Cast not supported yet!
- convertmodel
- netron
- 怎么生成有固定 shape 信息的模型？
- ncnnoptimize 怎么转成 fp16 模型
- ncnnoptimize 怎样查看模型的 FLOPS / 内存占用情况
- 怎么修改模型支持动态 shape？
- 如何将模型转换为代码内嵌到程序里？
- 如何加密模型？
- Linux下转的ncnn模型，Windows/MacOS/Android/.. 也能直接用吗？
- 如何去掉后处理，再导出 onnx？
- pytorch 有的层导不出 onnx 怎么办？
- 使用
-
- vkEnumeratePhysicalDevices failed -3
- vkCreateInstance failed -9
- ModuleNotFoundError: No module named 'ncnn.ncnn'
- fopen nanodet-m.param failed
- find_blob_index_by_name data / output / ... failed
- parse magic failed
- param is too old, please regenerate
- set_vulkan_compute failed, network use_vulkan_compute disabled
- 多个blob输入，多个blob输出，怎么做？
- Extractor extract 多次会重复计算吗？
- 如何看每一层的耗时？
- 如何转换 cv::Mat CV_8UC3 BGR 图片
- 如何转换 float 数据为 ncnn::Mat
- 如何初始化 ncnn::Mat 为全 0
- 如何转换 yuv 数据
- 如何 resize crop rotate 图片
- 如何人脸5点对齐
- 如何获得中间层的blob输出
- 为什么我使用GPU，但是GPU占用为0
- layer XYZ not exists or registered
- network graph not ready
- memory not 32-bit aligned at XYZ
- crash on android with '__kmp_abort_process'
- dlopen failed: library "libomp.so" not found
- crash when freeing a ncnn dynamic library(.dll/.so) built with openMP
- 跑出来的结果对不上
- 如何跑得更快？内存占用更少？库体积更小？
- 白嫖项目
- 其他
如何加入技术交流QQ群？
如何看作者b站直播？
nihui的bilibili直播间：https://live.bilibili.com/1264617
编译
怎样下载源码？
github
gitee
full-source.zip
怎么交叉编译？
cmake 工具链怎么设置啊？
The submodules were not downloaded! Please update submodules with "git submodule update --init" and try again
Could NOT find Protobuf (missing: Protobuf_INCLUDE_DIR)
sudo apt-get install libprotobuf-dev protobuf-compiler
Could NOT find CUDA (missing: CUDA_TOOLKIT_ROOT_DIR CUDA_INCLUDE_DIRS CUDA_CUDART_LIBRARY)
https://github.com/Tencent/ncnn/issues/1873
Could not find a package configuration file provided by "OpenCV" with any of the following names: OpenCVConfig.cmake opencv-config.cmake
set(OpenCV_DIR)
Could not find a package configuration file provided by "ncnn" with any of the following names: ncnnConfig.cmake ncnn-config.cmake
set(ncnn_DIR)
找不到 Vulkan
如何安装 vulkan sdk
cmake版本 3.10，否则没有带 FindVulkan.cmake
android-api >= 24
macos 要先执行安装脚本
undefined reference to __kmpc_for_static_init_4 __kmpc_for_static_fini __kmpc_fork_call ...
undefined reference to vkEnumerateInstanceExtensionProperties vkGetInstanceProcAddr vkQueueSubmit ...
undefined reference to glslang::InitializeProcess() glslang::TShader::TShader(EShLanguage) ...
undefined reference to AAssetManager_fromJava AAssetManager_open AAsset_seek ...
find_package(ncnn)
undefined reference to typeinfo for ncnn::Layer
opencv rtti -> opencv-mobile
undefined reference to __cpu_model
升级编译器 / libgcc_s libgcc
unrecognized command line option "-mavx2"
升级 gcc
为啥自己编译的ncnn android库特别大？
rtti/exceptions冲突
怎样添加ncnn库到项目中？cmake方式怎么用？
android
ios
linux
windows
macos
arm linux
转模型问题
caffe
./caffe2ncnn caffe.prototxt caffe.caffemodel ncnn.param ncnn.bin
mxnet
./mxnet2ncnn mxnet-symbol.json mxnet.params ncnn.param ncnn.bin
darknet
https://github.com/xiangweizeng/darknet2ncnn
pytorch - onnx
tensorflow 1.x/2.x - keras
https://github.com/MarsTechHAN/keras2ncnn @MarsTechHAN
tensorflow 2.x - mlir
通过MLIR将tensorflow2模型转换到ncnn @nihui
Shape not supported yet! Gather not supported yet! Cast not supported yet!
onnx-simplifier 静态shape
convertmodel
https://convertmodel.com/ @大老师
netron
https://github.com/lutzroeder/netron
怎么生成有固定 shape 信息的模型？
Input 0=w 1=h 2=c
why gpu能更快
ncnnoptimize 怎么转成 fp16 模型
ncnnoptimize model.param model.bin yolov5s-opt.param yolov5s-opt.bin 65536
ncnnoptimize 怎样查看模型的 FLOPS / 内存占用情况
怎么修改模型支持动态 shape？
Interp Reshape
如何将模型转换为代码内嵌到程序里？
ncnn2mem
如何加密模型？
https://zhuanlan.zhihu.com/p/268327784
Linux下转的ncnn模型，Windows/MacOS/Android/.. 也能直接用吗？
Yes，全平台通用
如何去掉后处理，再导出 onnx？
检测
参考up的一篇文章(https://zhuanlan.zhihu.com/p/128974102)，步骤三就是去掉后处理,再导出onnx,其中去掉后处理可以是项目内测试时去掉后续步骤的结果。
pytorch 有的层导不出 onnx 怎么办？
ONNX_ATEN_FALLBACK
完全自定义的op，先改成能导出的（如 concat slice），转到 ncnn 后再修改 param
使用
vkEnumeratePhysicalDevices failed -3
vkCreateInstance failed -9
驱动
opengl
ModuleNotFoundError: No module named 'ncnn.ncnn'
python setup.py develop
fopen nanodet-m.param failed
文件路径
working dir
File not found or not readable. Make sure that XYZ.param/XYZ.bin is accessible.
find_blob_index_by_name data / output / ... failed
layer name vs blob name
param.bin 应该用 xxx.id.h 的枚举
parse magic failed
param is too old, please regenerate
模型本身有问题
Your model file is being the old format converted by an old caffe2ncnn tool.
Checkout the latest ncnn code, build it and regenerate param and model binary files, and that should work.
Make sure that your param file starts with the magic number 7767517.
you may find more info on use-ncnn-with-alexnet
set_vulkan_compute failed, network use_vulkan_compute disabled
你应该在 load_param / load_model 之前设置 net.opt.use_vulkan_compute = true;
多个blob输入，多个blob输出，怎么做？
多次执行ex.input() 和 ex.extract()
ex.input();
ex.input();
ex.extract();
ex.extract();
Extractor extract 多次会重复计算吗？
如何看每一层的耗时？
cmake -DNCNN_BENCHMARK=ON ..
如何转换 cv::Mat CV_8UC3 BGR 图片
from_pixels to_pixels
如何转换 float 数据为 ncnn::Mat
https://github.com/Tencent/ncnn/wiki/use-ncnn-with-pytorch-or-onnx
如何初始化 ncnn::Mat 为全 0
mat.fill(0.f);
如何转换 yuv 数据
如何 resize crop rotate 图片
如何人脸5点对齐
get_affine_transform
warpaffine_bilinear_c3
如何获得中间层的blob输出
ncnn::Mat output;
ex.extract("your_blob_name", output);
为什么我使用GPU，但是GPU占用为0
windows 10 任务管理器 - 性能选项卡 - GPU - 选择其中一个视图左上角的下拉箭头切换到 Compute_0 / Compute_1 / Cuda
你还可以安装软件：GPU-Z
layer XYZ not exists or registered
Your network contains some operations that are not implemented in ncnn.
You may implement them as custom layer followed in how-to-implement-custom-layer-step-by-step.
Or you could simply register them as no-op if you are sure those operations make no sense.
class Noop : public ncnn::Layer {};
DEFINE_LAYER_CREATOR(Noop)
net.register_custom_layer("LinearRegressionOutput", Noop_layer_creator);
net.register_custom_layer("MAERegressionOutput", Noop_layer_creator);
network graph not ready
You shall call Net::load_param() first, then Net::load_model().
This error may also happens when Net::load_param() failed, but not properly handled.
For more information about the ncnn model load api, see ncnn-load-model
memory not 32-bit aligned at XYZ
The pointer passed to Net::load_param() or Net::load_model() is not 32bit aligned.
In practice, the head pointer of std::vector is not guaranteed to be 32bit aligned.
you can store your binary buffer in ncnn::Mat structure, its internal memory is aligned.
crash on android with '__kmp_abort_process'
This usually happens if you bundle multiple shared library with openmp linked
It is actually an issue of the android ndk https://github.com/android/ndk/issues/1028
On old android ndk, modify the link flags as
-Wl,-Bstatic -lomp -Wl,-Bdynamic
For recent ndk >= 21
-fstatic-openmp
dlopen failed: library "libomp.so" not found
Newer android ndk defaults to dynamic openmp runtime
modify the link flags as
-fstatic-openmp -fopenmp
crash when freeing a ncnn dynamic library(.dll/.so) built with openMP
for optimal performance, the openmp threadpool spin waits for about a second prior to shutting down in case more work becomes available.
If you unload a dynamic library that's in the process of spin-waiting, it will crash in the manner you see (most of the time).
Just set OMP_WAIT_POLICY=passive in your environment, before calling loadlibrary. or Just wait a few seconds before calling freelibrary.
You can also use the following method to set environment variables in your code:
for msvc++:
SetEnvironmentVariable(_T("OMP_WAIT_POLICY"), _T("passive"));
for g++:
setenv("OMP_WAIT_POLICY", "passive", 1)
reference: https://stackoverflow.com/questions/34439956/vc-crash-when-freeing-a-dll-built-with-openmp
跑出来的结果对不上
如何打印 ncnn::Mat 的值？
void pretty_print(const ncnn::Mat& m)
{
for (int q=0; q<m.c; q++)
{
const float* ptr = m.channel(q);
for (int y=0; y<m.h; y++)
{
for (int x=0; x<m.w; x++)
{
printf("%f ", ptr[x]);
}
ptr += m.w;
printf("\n");
}
printf("------------------------\n");
}
}
In Android Studio, printf will not work, you can use __android_log_print instead. Example :
#include <android/log.h> // Don't forget this
void pretty_print(const ncnn::Mat& m)
{
for (int q=0; q<m.c; q++)
{
for (int y=0; y<m.h; y++)
{
for (int x=0; x<m.w; x++)
{
__android_log_print(ANDROID_LOG_DEBUG,"LOG_TAG","ncnn Mat is : %f", m.channel(q).row(y)[x]);
}
}
}
}
如何可视化 ncnn::Mat 的值？
void visualize(const char* title, const ncnn::Mat& m)
{
std::vector<cv::Mat> normed_feats(m.c);
for (int i=0; i<m.c; i++)
{
cv::Mat tmp(m.h, m.w, CV_32FC1, (void*)(const float*)m.channel(i));
cv::normalize(tmp, normed_feats[i], 0, 255, cv::NORM_MINMAX, CV_8U);
cv::cvtColor(normed_feats[i], normed_feats[i], cv::COLOR_GRAY2BGR);
// check NaN
for (int y=0; y<m.h; y++)
{
const float* tp = tmp.ptr<float>(y);
uchar* sp = normed_feats[i].ptr<uchar>(y);
for (int x=0; x<m.w; x++)
{
float v = tp[x];
if (v != v)
{
sp[0] = 0;
sp[1] = 0;
sp[2] = 255;
}
sp += 3;
}
}
}
int tw = m.w < 10 ? 32 : m.w < 20 ? 16 : m.w < 40 ? 8 : m.w < 80 ? 4 : m.w < 160 ? 2 : 1;
int th = (m.c - 1) / tw + 1;
cv::Mat show_map(m.h * th, m.w * tw, CV_8UC3);
show_map = cv::Scalar(127);
// tile
for (int i=0; i<m.c; i++)
{
int ty = i / tw;
int tx = i % tw;
normed_feats[i].copyTo(show_map(cv::Rect(tx * m.w, ty * m.h, m.w, m.h)));
}
cv::resize(show_map, show_map, cv::Size(0,0), 2, 2, cv::INTER_NEAREST);
cv::imshow(title, show_map);
}
总是输出第一张图的结果
复用 Extractor？！
启用fp16时的精度差异
net.opt.use_fp16_packed = false;
net.opt.use_fp16_storage = false;
net.opt.use_fp16_arithmetic = false;
https://github.com/Tencent/ncnn/wiki/FAQ-ncnn-produce-wrong-result
如何跑得更快？内存占用更少？库体积更小？
fp32 fp16
大小核绑定
get_cpu_count get_gpu_count
ncnnoptimize
如何使用量化工具？
Post Training Quantization Tools
如何设置线程数？
如何降低CPU占用率？
net.opt.openmp_blocktime = 0;
OMP_WAIT_POLICY=passive
如何 batch inference？
partial graph inference
先 extract 分类，判断后，再 extract bbox
如何启用 bf16s 加速？
net.opt.use_packing_layout = true;
net.opt.use_bf16_storage = true;
A53
如何裁剪更小的 ncnn 库？
https://github.com/Tencent/ncnn/wiki/build-minimal-library
net.opt sgemm winograd fp16_storage 各是有什么作用？
对内存消耗的影响
白嫖项目
nanodet
其他
up主用的什么系统/编辑器/开发环境？
| 软件类型 | 软件名称 |
|---|---|
| 系统 | Fedora |
| 桌面环境 | KDE |
| 编辑器 | Kate |
| 画草图 | kolourpaint |
| 画函数图像 | kmplot |
| bilibili直播 | OBS |