diff --git a/model_zoo/official/cv/deeptext/README.md b/model_zoo/official/cv/deeptext/README.md index 77fefad87d..a5178fe679 100644 --- a/model_zoo/official/cv/deeptext/README.md +++ b/model_zoo/official/cv/deeptext/README.md @@ -64,9 +64,11 @@ Here we used 4 datasets for training, and 1 datasets for Evaluation. . └─deeptext ├─README.md + ├─ascend310_infer #application for 310 inference ├─scripts ├─run_standalone_train_ascend.sh # launch standalone training with ascend platform(1p) ├─run_distribute_train_ascend.sh # launch distributed training with ascend platform(8p) + ├─run_infer_310.sh # shell script for 310 inference └─run_eval_ascend.sh # launch evaluating with ascend platform ├─src ├─DeepText @@ -81,12 +83,14 @@ Here we used 4 datasets for training, and 1 datasets for Evaluation. ├─rpn.py # region-proposal network └─vgg16.py # backbone ├─config.py # training configuration + ├─aipp.cfg # aipp config file ├─dataset.py # data proprocessing ├─lr_schedule.py # learning rate scheduler ├─network_define.py # network definition └─utils.py # some functions which is commonly used ├─eval.py # eval net ├─export.py # export checkpoint, surpport .onnx, .air, .mindir convert + ├─postprogress.py # post process for 310 inference └─train.py # train net ``` @@ -168,6 +172,35 @@ Evaluation result will be stored in the example path, you can find result like t class 1 precision is 88.01%, recall is 82.77% ``` +## Model Export + +```shell +python export.py --ckpt_file [CKPT_PATH] --device_target [DEVICE_TARGET] --file_format[EXPORT_FORMAT] +``` + +`EXPORT_FORMAT` should be in ["AIR", "MINDIR"] + +## Inference Process + +### Usage + +Before performing inference, the air file must bu exported by export script on the Ascend910 environment. + +```shell +# Ascend310 inference +bash run_infer_310.sh [MINDIR_PATH] [DATA_PATH] [LABEL_PATH] [DEVICE_ID] +``` + +### result + +Inference result is saved in current path, you can find result like this in acc.log file. + +```python +======================================== + +class 1 precision is 84.24%, recall is 87.40%, F1 is 85.79% +``` + # [Model description](#contents) ## [Performance](#contents) @@ -177,7 +210,7 @@ class 1 precision is 88.01%, recall is 82.77% | Parameters | Ascend | | -------------------------- | ------------------------------------------------------------ | | Model Version | Deeptext | -| Resource | Ascend 910, cpu:2.60GHz 192cores, memory:755G | +| Resource | Ascend 910, cpu:2.60GHz 192cores, memory:755G, OS:Euler2.8 | | uploaded Date | 12/26/2020 | | MindSpore Version | 1.1.0 | | Dataset | 66040 images | diff --git a/model_zoo/official/cv/deeptext/ascend310_infer/CMakeLists.txt b/model_zoo/official/cv/deeptext/ascend310_infer/CMakeLists.txt new file mode 100644 index 0000000000..ee3c854473 --- /dev/null +++ b/model_zoo/official/cv/deeptext/ascend310_infer/CMakeLists.txt @@ -0,0 +1,14 @@ +cmake_minimum_required(VERSION 3.14.1) +project(Ascend310Infer) +add_compile_definitions(_GLIBCXX_USE_CXX11_ABI=0) +set(CMAKE_CXX_FLAGS "${CMAKE_CXX_FLAGS} -O0 -g -std=c++17 -Werror -Wall -fPIE -Wl,--allow-shlib-undefined") +set(PROJECT_SRC_ROOT ${CMAKE_CURRENT_LIST_DIR}/) +option(MINDSPORE_PATH "mindspore install path" "") +include_directories(${MINDSPORE_PATH}) +include_directories(${MINDSPORE_PATH}/include) +include_directories(${PROJECT_SRC_ROOT}) +find_library(MS_LIB libmindspore.so ${MINDSPORE_PATH}/lib) +file(GLOB_RECURSE MD_LIB ${MINDSPORE_PATH}/_c_dataengine*) + +add_executable(main src/main.cc src/utils.cc) +target_link_libraries(main ${MS_LIB} ${MD_LIB} gflags) diff --git a/model_zoo/official/cv/deeptext/ascend310_infer/build.sh b/model_zoo/official/cv/deeptext/ascend310_infer/build.sh new file mode 100644 index 0000000000..8bf761bc33 --- /dev/null +++ b/model_zoo/official/cv/deeptext/ascend310_infer/build.sh @@ -0,0 +1,23 @@ +#!/bin/bash +# Copyright 2021 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. +# ============================================================================ + +if [ ! -d out ]; then + mkdir out +fi +cd out +cmake .. \ + -DMINDSPORE_PATH="`pip show mindspore-ascend | grep Location | awk '{print $2"/mindspore"}' | xargs realpath`" +make diff --git a/model_zoo/official/cv/deeptext/ascend310_infer/inc/utils.h b/model_zoo/official/cv/deeptext/ascend310_infer/inc/utils.h new file mode 100644 index 0000000000..efebe03a8c --- /dev/null +++ b/model_zoo/official/cv/deeptext/ascend310_infer/inc/utils.h @@ -0,0 +1,32 @@ +/** + * Copyright 2021 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_INFERENCE_UTILS_H_ +#define MINDSPORE_INFERENCE_UTILS_H_ + +#include +#include +#include +#include +#include +#include "include/api/types.h" + +std::vector GetAllFiles(std::string_view dirName); +DIR *OpenDir(std::string_view dirName); +std::string RealPath(std::string_view path); +mindspore::MSTensor ReadFileToTensor(const std::string &file); +int WriteResult(const std::string& imageFile, const std::vector &outputs); +#endif diff --git a/model_zoo/official/cv/deeptext/ascend310_infer/src/main.cc b/model_zoo/official/cv/deeptext/ascend310_infer/src/main.cc new file mode 100644 index 0000000000..93854ee1f6 --- /dev/null +++ b/model_zoo/official/cv/deeptext/ascend310_infer/src/main.cc @@ -0,0 +1,154 @@ +/** + * Copyright 2021 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. + */ + +#include +#include +#include +#include +#include +#include +#include +#include +#include +#include + +#include "../inc/utils.h" +#include "minddata/dataset/include/execute.h" +#include "minddata/dataset/include/transforms.h" +#include "minddata/dataset/include/vision.h" +#include "minddata/dataset/include/vision_ascend.h" +#include "include/api/types.h" +#include "include/api/model.h" +#include "include/api/serialization.h" +#include "include/api/context.h" + + +using mindspore::Context; +using mindspore::Serialization; +using mindspore::Model; +using mindspore::Status; +using mindspore::dataset::Execute; +using mindspore::MSTensor; +using mindspore::ModelType; +using mindspore::GraphCell; +using mindspore::kSuccess; +using mindspore::Graph; +using mindspore::dataset::vision::DvppDecodeResizeJpeg; + +DEFINE_string(model_path, "", "model path"); +DEFINE_string(dataset_path, ".", "dataset path"); +DEFINE_int32(input_width, 960, "input width"); +DEFINE_int32(input_height, 576, "inputheight"); +DEFINE_int32(device_id, 0, "device id"); +DEFINE_string(precision_mode, "allow_fp32_to_fp16", "precision mode"); +DEFINE_string(op_select_impl_mode, "", "op select impl mode"); +DEFINE_string(aipp_path, "./aipp.cfg", "aipp path"); +DEFINE_string(dump_config_file, "./acl.json", "dump config file"); +DEFINE_string(device_target, "Ascend310", "device target"); + +int main(int argc, char **argv) { + gflags::ParseCommandLineFlags(&argc, &argv, true); + if (RealPath(FLAGS_model_path).empty()) { + std::cout << "Invalid mindir" << std::endl; + return 1; + } + if (RealPath(FLAGS_aipp_path).empty()) { + std::cout << "Invalid aipp path" << std::endl; + return 1; + } + + auto context = std::make_shared(); + auto ascend310_info = std::make_shared(); + ascend310_info->SetDeviceID(FLAGS_device_id); + ascend310_info->SetInsertOpConfigPath(FLAGS_aipp_path); + context->MutableDeviceInfo().push_back(ascend310_info); + + Graph graph; + Status ret = Serialization::Load(FLAGS_model_path, ModelType::kMindIR, &graph); + if (ret != kSuccess) { + std::cout << "Load model failed." << std::endl; + return 1; + } + + Model model; + ret = model.Build(GraphCell(graph), context); + if (ret != kSuccess) { + std::cout << "ERROR: Build failed." << std::endl; + return 1; + } + + std::vector modelInputs = model.GetInputs(); + + auto all_files = GetAllFiles(FLAGS_dataset_path); + if (all_files.empty()) { + std::cout << "ERROR: no input data." << std::endl; + return 1; + } + + std::map costTime_map; + size_t size = all_files.size(); + + Execute transform(std::shared_ptr(new DvppDecodeResizeJpeg({576, 960}))); + + for (size_t i = 0; i < size; ++i) { + struct timeval start; + struct timeval end; + double startTime_ms; + double endTime_ms; + std::vector inputs; + std::vector outputs; + + std::cout << "Start predict input files:" << all_files[i] << std::endl; + + mindspore::MSTensor imgDvpp; + transform(ReadFileToTensor(all_files[i]), &imgDvpp); + + inputs.emplace_back(modelInputs[0].Name(), modelInputs[0].DataType(), modelInputs[0].Shape(), + imgDvpp.Data().get(), imgDvpp.DataSize()); + + gettimeofday(&start, NULL); + ret = model.Predict(inputs, &outputs); + gettimeofday(&end, NULL); + if (ret != kSuccess) { + std::cout << "Predict " << all_files[i] << " failed." << std::endl; + return 1; + } + startTime_ms = (1.0 * start.tv_sec * 1000000 + start.tv_usec) / 1000; + endTime_ms = (1.0 * end.tv_sec * 1000000 + end.tv_usec) / 1000; + costTime_map.insert(std::pair(startTime_ms, endTime_ms)); + WriteResult(all_files[i], outputs); + } + double average = 0.0; + int infer_cnt = 0; + + for (auto iter = costTime_map.begin(); iter != costTime_map.end(); iter++) { + double diff = 0.0; + diff = iter->second - iter->first; + average += diff; + infer_cnt++; + } + + average = average/infer_cnt; + std::stringstream timeCost; + timeCost << "NN inference cost average time: "<< average << "ms of infer_count " << infer_cnt << std::endl; + std::cout << "NN inference cost average time: "<< average << "ms of infer_count " << infer_cnt << std::endl; + std::string file_name = "./time_Result" + std::string("/test_perform_static.txt"); + std::ofstream file_stream(file_name.c_str(), std::ios::trunc); + file_stream << timeCost.str(); + file_stream.close(); + costTime_map.clear(); + return 0; +} diff --git a/model_zoo/official/cv/deeptext/ascend310_infer/src/utils.cc b/model_zoo/official/cv/deeptext/ascend310_infer/src/utils.cc new file mode 100644 index 0000000000..b509c57f82 --- /dev/null +++ b/model_zoo/official/cv/deeptext/ascend310_infer/src/utils.cc @@ -0,0 +1,130 @@ +/** + * Copyright 2021 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. + */ + +#include "inc/utils.h" + +#include +#include +#include + +using mindspore::MSTensor; +using mindspore::DataType; + +std::vector GetAllFiles(std::string_view dirName) { + struct dirent *filename; + DIR *dir = OpenDir(dirName); + if (dir == nullptr) { + return {}; + } + std::vector res; + while ((filename = readdir(dir)) != nullptr) { + std::string dName = std::string(filename->d_name); + if (dName == "." || dName == ".." || filename->d_type != DT_REG) { + continue; + } + res.emplace_back(std::string(dirName) + "/" + filename->d_name); + } + std::sort(res.begin(), res.end()); + for (auto &f : res) { + std::cout << "image file: " << f << std::endl; + } + return res; +} + +int WriteResult(const std::string& imageFile, const std::vector &outputs) { + std::string homePath = "./result_Files"; + for (size_t i = 0; i < outputs.size(); ++i) { + size_t outputSize; + std::shared_ptr netOutput; + netOutput = outputs[i].Data(); + outputSize = outputs[i].DataSize(); + int pos = imageFile.rfind('/'); + std::string fileName(imageFile, pos + 1); + fileName.replace(fileName.find('.'), fileName.size() - fileName.find('.'), '_' + std::to_string(i) + ".bin"); + std::string outFileName = homePath + "/" + fileName; + FILE * outputFile = fopen(outFileName.c_str(), "wb"); + fwrite(netOutput.get(), outputSize, sizeof(char), outputFile); + fclose(outputFile); + outputFile = nullptr; + } + return 0; +} + +mindspore::MSTensor ReadFileToTensor(const std::string &file) { + if (file.empty()) { + std::cout << "Pointer file is nullptr" << std::endl; + return mindspore::MSTensor(); + } + + std::ifstream ifs(file); + if (!ifs.good()) { + std::cout << "File: " << file << " is not exist" << std::endl; + return mindspore::MSTensor(); + } + + if (!ifs.is_open()) { + std::cout << "File: " << file << "open failed" << std::endl; + return mindspore::MSTensor(); + } + + ifs.seekg(0, std::ios::end); + size_t size = ifs.tellg(); + mindspore::MSTensor buffer(file, mindspore::DataType::kNumberTypeUInt8, {static_cast(size)}, nullptr, size); + + ifs.seekg(0, std::ios::beg); + ifs.read(reinterpret_cast(buffer.MutableData()), size); + ifs.close(); + + return buffer; +} + + +DIR *OpenDir(std::string_view dirName) { + if (dirName.empty()) { + std::cout << " dirName is null ! " << std::endl; + return nullptr; + } + std::string realPath = RealPath(dirName); + struct stat s; + lstat(realPath.c_str(), &s); + if (!S_ISDIR(s.st_mode)) { + std::cout << "dirName is not a valid directory !" << std::endl; + return nullptr; + } + DIR *dir; + dir = opendir(realPath.c_str()); + if (dir == nullptr) { + std::cout << "Can not open dir " << dirName << std::endl; + return nullptr; + } + std::cout << "Successfully opened the dir " << dirName << std::endl; + return dir; +} + +std::string RealPath(std::string_view path) { + char realPathMem[PATH_MAX] = {0}; + char *realPathRet = nullptr; + realPathRet = realpath(path.data(), realPathMem); + + if (realPathRet == nullptr) { + std::cout << "File: " << path << " is not exist."; + return ""; + } + + std::string realPath(realPathMem); + std::cout << path << " realpath is: " << realPath << std::endl; + return realPath; +} diff --git a/model_zoo/official/cv/deeptext/export.py b/model_zoo/official/cv/deeptext/export.py new file mode 100644 index 0000000000..acbf3d7229 --- /dev/null +++ b/model_zoo/official/cv/deeptext/export.py @@ -0,0 +1,53 @@ +# Copyright 2021 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. +# ============================================================================ +"""export checkpoint file into air, mindir models""" +import argparse +import numpy as np + +import mindspore as ms +from mindspore import Tensor, load_checkpoint, load_param_into_net, export, context + +from src.Deeptext.deeptext_vgg16 import Deeptext_VGG16_Infer +from src.config import config + +parser = argparse.ArgumentParser(description='deeptext export') +parser.add_argument("--device_id", type=int, default=0, help="Device id") +parser.add_argument("--batch_size", type=int, default=1, help="batch size") +parser.add_argument("--file_name", type=str, default="deeptext", help="output file name.") +parser.add_argument("--file_format", type=str, choices=["AIR", "MINDIR"], default="MINDIR", help="file format") +parser.add_argument("--device_target", type=str, choices=["Ascend", "GPU", "CPU"], default="Ascend", + help="device target") +parser.add_argument('--ckpt_file', type=str, default='', help='deeptext ckpt file.') +args = parser.parse_args() + +config.test_batch_size = args.batch_size +context.set_context(mode=context.GRAPH_MODE, device_target=args.device_target) +context.set_context(device_id=args.device_id) + +if __name__ == '__main__': + net = Deeptext_VGG16_Infer(config=config) + net.set_train(False) + + param_dict = load_checkpoint(args.ckpt_file) + + param_dict_new = {} + for key, value in param_dict.items(): + param_dict_new["network." + key] = value + + load_param_into_net(net, param_dict_new) + + img_data = Tensor(np.zeros([config.test_batch_size, 3, config.img_height, config.img_width]), ms.float16) + + export(net, img_data, file_name=args.file_name, file_format=args.file_format) diff --git a/model_zoo/official/cv/deeptext/postprocess.py b/model_zoo/official/cv/deeptext/postprocess.py new file mode 100644 index 0000000000..273131791a --- /dev/null +++ b/model_zoo/official/cv/deeptext/postprocess.py @@ -0,0 +1,123 @@ +# Copyright 2021 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. +# ============================================================================ + +"""Evaluation for Deeptext""" +import argparse +import os + +import numpy as np +from src.config import config +from src.utils import metrics +from PIL import Image +import mmcv + +parser = argparse.ArgumentParser(description="Deeptext evaluation") +parser.add_argument("--result_path", type=str, required=True, help="result file path") +parser.add_argument("--label_path", type=str, required=True, help="label path") +parser.add_argument("--img_path", type=str, required=True, help="img path") +args_opt = parser.parse_args() + +config.test_batch_size = 1 + +def get_pred(file, result_path): + file_name = file.split('.')[0][3:] + all_bbox_file = os.path.join(result_path, file_name + "_0.bin") + all_label_file = os.path.join(result_path, file_name + "_1.bin") + all_mask_file = os.path.join(result_path, file_name + "_2.bin") + all_bbox = np.fromfile(all_bbox_file, dtype=np.float16).reshape(config.test_batch_size, 1000, 5) + all_label = np.fromfile(all_label_file, dtype=np.int32).reshape(config.test_batch_size, 1000, 1) + all_mask = np.fromfile(all_mask_file, dtype=np.bool).reshape(config.test_batch_size, 1000, 1) + + return all_bbox, all_label, all_mask + +def get_gt_bboxes_labels(label_file, img_file): + img_data = np.array(Image.open(img_file)) + img_data, w_scale, h_scale = mmcv.imresize( + img_data, (config.img_width, config.img_height), return_scale=True) + scale_factor = np.array( + [w_scale, h_scale, w_scale, h_scale], dtype=np.float32) + img_shape = (config.img_height, config.img_width, 1.0) + img_shape = np.asarray(img_shape, dtype=np.float32) + + file = open(label_file) + lines = file.readlines() + boxes = [] + gt_label = [] + for line in lines: + label_info = line.split(",") + boxes.append([float(label_info[0]), float(label_info[1]), float(label_info[2]), float(label_info[3])]) + gt_label.append(int(1)) + + gt_bboxes = np.array(boxes) + gt_bboxes = gt_bboxes * scale_factor + + gt_bboxes[:, 0::2] = np.clip(gt_bboxes[:, 0::2], 0, img_shape[1] - 1) + gt_bboxes[:, 1::2] = np.clip(gt_bboxes[:, 1::2], 0, img_shape[0] - 1) + + return gt_bboxes, gt_label + +def deeptext_eval_test(result_path='', label_path='', img_path=''): + eval_iter = 0 + + print("\n========================================\n") + print("Processing, please wait a moment.") + max_num = 32 + + pred_data = [] + files = os.listdir(label_path) + for file in files: + eval_iter = eval_iter + 1 + img_file = os.path.join(img_path, file.split('gt_')[1].replace("txt", "jpg")) + + label_file = os.path.join(label_path, file) + gt_bboxes, gt_labels = get_gt_bboxes_labels(label_file, img_file) + + gt_bboxes = np.array(gt_bboxes).astype(np.float32) + + all_bbox, all_label, all_mask = get_pred(file, result_path) + all_label = all_label + 1 + + + for j in range(config.test_batch_size): + all_bbox_squee = np.squeeze(all_bbox[j, :, :]) + all_label_squee = np.squeeze(all_label[j, :, :]) + all_mask_squee = np.squeeze(all_mask[j, :, :]) + + all_bboxes_tmp_mask = all_bbox_squee[all_mask_squee, :] + all_labels_tmp_mask = all_label_squee[all_mask_squee] + + if all_bboxes_tmp_mask.shape[0] > max_num: + inds = np.argsort(-all_bboxes_tmp_mask[:, -1]) + inds = inds[:max_num] + all_bboxes_tmp_mask = all_bboxes_tmp_mask[inds] + all_labels_tmp_mask = all_labels_tmp_mask[inds] + + pred_data.append({"boxes": all_bboxes_tmp_mask, + "labels": all_labels_tmp_mask, + "gt_bboxes": gt_bboxes, + "gt_labels": gt_labels}) + + precisions, recalls = metrics(pred_data) + print("\n========================================\n") + for i in range(config.num_classes - 1): + j = i + 1 + f1 = (2 * precisions[j] * recalls[j]) / (precisions[j] + recalls[j] + 1e-6) + print("class {} precision is {:.2f}%, recall is {:.2f}%," + "F1 is {:.2f}%".format(j, precisions[j] * 100, recalls[j] * 100, f1 * 100)) + if config.use_ambigous_sample: + break + +if __name__ == '__main__': + deeptext_eval_test(args_opt.result_path, args_opt.label_path, args_opt.img_path) diff --git a/model_zoo/official/cv/deeptext/scripts/run_infer_310.sh b/model_zoo/official/cv/deeptext/scripts/run_infer_310.sh new file mode 100755 index 0000000000..5e9ea2ae4e --- /dev/null +++ b/model_zoo/official/cv/deeptext/scripts/run_infer_310.sh @@ -0,0 +1,107 @@ +#!/bin/bash +# Copyright 2021 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. +# ============================================================================ + +if [[ $# -lt 3 || $# -gt 4 ]]; then + echo "Usage: sh run_infer_310.sh [MINDIR_PATH] [DATA_PATH] [LABEL_PATH] [DEVICE_ID] + DEVICE_ID is optional, it can be set by environment variable device_id, otherwise the value is zero" +exit 1 +fi + +get_real_path(){ + if [ "${1:0:1}" == "/" ]; then + echo "$1" + else + echo "$(realpath -m $PWD/$1)" + fi +} + +model=$(get_real_path $1) +data_path=$(get_real_path $2) +label_path=$(get_real_path $3) +if [ $# == 4 ]; then + device_id=$4 +elif [ $# == 3 ]; then + if [ -z $device_id ]; then + device_id=0 + else + device_id=$device_id + fi +fi + +echo $model +echo $data_path +echo $label_path +echo $device_id + +export ASCEND_HOME=/usr/local/Ascend/ +if [ -d ${ASCEND_HOME}/ascend-toolkit ]; then + export PATH=$ASCEND_HOME/ascend-toolkit/latest/fwkacllib/ccec_compiler/bin:$ASCEND_HOME/ascend-toolkit/latest/atc/bin:$PATH + export LD_LIBRARY_PATH=/usr/local/lib:$ASCEND_HOME/ascend-toolkit/latest/atc/lib64:$ASCEND_HOME/ascend-toolkit/latest/fwkacllib/lib64:$ASCEND_HOME/driver/lib64:$ASCEND_HOME/add-ons:$LD_LIBRARY_PATH + export TBE_IMPL_PATH=$ASCEND_HOME/ascend-toolkit/latest/opp/op_impl/built-in/ai_core/tbe + export PYTHONPATH=${TBE_IMPL_PATH}:$ASCEND_HOME/ascend-toolkit/latest/fwkacllib/python/site-packages:$PYTHONPATH + export ASCEND_OPP_PATH=$ASCEND_HOME/ascend-toolkit/latest/opp +else + export PATH=$ASCEND_HOME/atc/ccec_compiler/bin:$ASCEND_HOME/atc/bin:$PATH + export LD_LIBRARY_PATH=/usr/local/lib:$ASCEND_HOME/atc/lib64:$ASCEND_HOME/acllib/lib64:$ASCEND_HOME/driver/lib64:$ASCEND_HOME/add-ons:$LD_LIBRARY_PATH + export PYTHONPATH=$ASCEND_HOME/atc/python/site-packages:$PYTHONPATH + export ASCEND_OPP_PATH=$ASCEND_HOME/opp +fi + +function compile_app() +{ + cd ../ascend310_infer + if [ -f "Makefile" ]; then + make clean + fi + sh build.sh &> build.log + + if [ $? -ne 0 ]; then + echo "compile app code failed" + exit 1 + fi + cd - +} + +function infer() +{ + if [ -d result_Files ]; then + rm -rf ./result_Files + fi + if [ -d time_Result ]; then + rm -rf ./time_Result + fi + mkdir result_Files + mkdir time_Result + ../ascend310_infer/out/main --model_path=$model --dataset_path=$data_path --device_id=$device_id --aipp_path ../src/aipp.cfg &> infer.log + + if [ $? -ne 0 ]; then + echo "execute inference failed" + exit 1 + fi +} + +function cal_acc() +{ + python ../postprocess.py --label_path=$label_path --result_path=result_Files --img_path=$data_path &> acc.log & + if [ $? -ne 0 ]; then + echo "calculate accuracy failed" + exit 1 + fi +} + +compile_app +infer +cal_acc diff --git a/model_zoo/official/cv/deeptext/src/Deeptext/deeptext_vgg16.py b/model_zoo/official/cv/deeptext/src/Deeptext/deeptext_vgg16.py index 9bfdafeab4..de6c3871f4 100644 --- a/model_zoo/official/cv/deeptext/src/Deeptext/deeptext_vgg16.py +++ b/model_zoo/official/cv/deeptext/src/Deeptext/deeptext_vgg16.py @@ -430,3 +430,13 @@ class Deeptext_VGG16(nn.Cell): multi_level_anchors += (Tensor(anchors.astype(np.float32)),) return multi_level_anchors + +class Deeptext_VGG16_Infer(nn.Cell): + def __init__(self, config): + super(Deeptext_VGG16_Infer, self).__init__() + self.network = Deeptext_VGG16(config) + self.network.set_train(False) + + def construct(self, img_data): + output = self.network(img_data, None, None, None, None) + return output diff --git a/model_zoo/official/cv/deeptext/src/aipp.cfg b/model_zoo/official/cv/deeptext/src/aipp.cfg new file mode 100644 index 0000000000..363d5d36fd --- /dev/null +++ b/model_zoo/official/cv/deeptext/src/aipp.cfg @@ -0,0 +1,26 @@ +aipp_op { + aipp_mode : static + input_format : YUV420SP_U8 + related_input_rank : 0 + csc_switch : true + rbuv_swap_switch : false + matrix_r0c0 : 256 + matrix_r0c1 : 0 + matrix_r0c2 : 359 + matrix_r1c0 : 256 + matrix_r1c1 : -88 + matrix_r1c2 : -183 + matrix_r2c0 : 256 + matrix_r2c1 : 454 + matrix_r2c2 : 0 + input_bias_0 : 0 + input_bias_1 : 128 + input_bias_2 : 128 + + mean_chn_0 : 124 + mean_chn_1 : 117 + mean_chn_2 : 104 + var_reci_chn_0 : 0.0171247538316637 + var_reci_chn_1 : 0.0175070028011204 + var_reci_chn_2 : 0.0174291938997821 +} \ No newline at end of file