Merge remote-tracking branch 'upstream/dygraph' into dy1

README.md

@@ -122,8 +122,7 @@ For a new language request, please refer to [Guideline for new language_requests
     <img src="./doc/ppocr_framework.png" width="800">
 </div>
 
-PP-OCR is a practical ultra-lightweight OCR system. It is mainly composed of three parts: DB text detection, detection frame correction and CRNN text recognition. The system adopts 19 effective strategies from 8 aspects including backbone network selection and adjustment, prediction head design, data augmentation, learning rate transformation strategy, regularization parameter selection, pre-training model use, and automatic model tailoring and quantization to optimize and slim down the models of each module. The final results are an ultra-lightweight Chinese and English OCR model with an overall size of 3.5M and a 2.8M English digital OCR model. For more details, please refer to the PP-OCR technical article (https://arxiv.org/abs/2009.09941). Besides, The implementation of the FPGM Pruner and PACT quantization is based on [PaddleSlim](https://github.com/PaddlePaddle/PaddleSlim).
-
+PP-OCR is a practical ultra-lightweight OCR system. It is mainly composed of three parts: DB text detection[2], detection frame correction and CRNN text recognition[7]. The system adopts 19 effective strategies from 8 aspects including backbone network selection and adjustment, prediction head design, data augmentation, learning rate transformation strategy, regularization parameter selection, pre-training model use, and automatic model tailoring and quantization to optimize and slim down the models of each module. The final results are an ultra-lightweight Chinese and English OCR model with an overall size of 3.5M and a 2.8M English digital OCR model. For more details, please refer to the PP-OCR technical article (https://arxiv.org/abs/2009.09941). Besides, The implementation of the FPGM Pruner [8] and PACT quantization [9] is based on [PaddleSlim](https://github.com/PaddlePaddle/PaddleSlim).
 
 
 ## Visualization [more](./doc/doc_en/visualization_en.md)

README_ch.md

@@ -8,8 +8,8 @@ PaddleOCR同时支持动态图与静态图两种编程范式
 - 静态图版本：develop分支
 
 **近期更新**
+- 2021.1.4 [FAQ](./doc/doc_ch/FAQ.md)新增5个高频问题，总数142个，每周一都会更新，欢迎大家持续关注。
 - 2020.12.15 更新数据合成工具[Style-Text](./StyleText/README_ch.md)，可以批量合成大量与目标场景类似的图像，在多个场景验证，效果明显提升。
-- 2020.12.14 [FAQ](./doc/doc_ch/FAQ.md)新增5个高频问题，总数127个，每周一都会更新，欢迎大家持续关注。
 - 2020.11.25 更新半自动标注工具[PPOCRLabel](./PPOCRLabel/README_ch.md)，辅助开发者高效完成标注任务，输出格式与PP-OCR训练任务完美衔接。
 - 2020.9.22 更新PP-OCR技术文章，https://arxiv.org/abs/2009.09941
 - [More](./doc/doc_ch/update.md)
@@ -101,8 +101,8 @@ PaddleOCR同时支持动态图与静态图两种编程范式
 - [效果展示](#效果展示)
 - FAQ
     - [【精选】OCR精选10个问题](./doc/doc_ch/FAQ.md)
-    - [【理论篇】OCR通用30个问题](./doc/doc_ch/FAQ.md)
-    - [【实战篇】PaddleOCR实战84个问题](./doc/doc_ch/FAQ.md)
+    - [【理论篇】OCR通用31个问题](./doc/doc_ch/FAQ.md)
+    - [【实战篇】PaddleOCR实战101个问题](./doc/doc_ch/FAQ.md)
 - [技术交流群](#欢迎加入PaddleOCR技术交流群)
 - [参考文献](./doc/doc_ch/reference.md)
 - [许可证书](#许可证书)
@@ -115,7 +115,7 @@ PaddleOCR同时支持动态图与静态图两种编程范式
     <img src="./doc/ppocr_framework.png" width="800">
 </div>
 
-PP-OCR是一个实用的超轻量OCR系统。主要由DB文本检测、检测框矫正和CRNN文本识别三部分组成。该系统从骨干网络选择和调整、预测头部的设计、数据增强、学习率变换策略、正则化参数选择、预训练模型使用以及模型自动裁剪量化8个方面，采用19个有效策略，对各个模块的模型进行效果调优和瘦身，最终得到整体大小为3.5M的超轻量中英文OCR和2.8M的英文数字OCR。更多细节请参考PP-OCR技术方案 https://arxiv.org/abs/2009.09941 。其中FPGM裁剪器和PACT量化的实现可以参考[PaddleSlim](https://github.com/PaddlePaddle/PaddleSlim)。
+PP-OCR是一个实用的超轻量OCR系统。主要由DB文本检测[2]、检测框矫正和CRNN文本识别三部分组成[7]。该系统从骨干网络选择和调整、预测头部的设计、数据增强、学习率变换策略、正则化参数选择、预训练模型使用以及模型自动裁剪量化8个方面，采用19个有效策略，对各个模块的模型进行效果调优和瘦身，最终得到整体大小为3.5M的超轻量中英文OCR和2.8M的英文数字OCR。更多细节请参考PP-OCR技术方案 https://arxiv.org/abs/2009.09941 。其中FPGM裁剪器[8]和PACT量化[9]的实现可以参考[PaddleSlim](https://github.com/PaddlePaddle/PaddleSlim)。
 
 <a name="效果展示"></a>
 ## 效果展示 [more](./doc/doc_ch/visualization.md)

StyleText/README.md

@@ -22,7 +22,7 @@ English | [简体中文](README_ch.md)
 </div>
 
 
-The Style-Text data synthesis tool is a tool based on Baidu's self-developed text editing algorithm "Editing Text in the Wild" [https://arxiv.org/abs/1908.03047](https://arxiv.org/abs/1908.03047).
+The Style-Text data synthesis tool is a tool based on Baidu and HUST cooperation research work, "Editing Text in the Wild" [https://arxiv.org/abs/1908.03047](https://arxiv.org/abs/1908.03047).
 
 Different from the commonly used GAN-based data synthesis tools, the main framework of Style-Text includes:
 * (1) Text foreground style transfer module.
@@ -69,12 +69,14 @@ fusion_generator:
 1. You can run `tools/synth_image` and generate the demo image, which is saved in the current folder.
 
 ```python
-python3 -m tools.synth_image -c configs/config.yml --style_image examples/style_images/2.jpg --text_corpus PaddleOCR --language en
+python3 tools/synth_image.py -c configs/config.yml --style_image examples/style_images/2.jpg --text_corpus PaddleOCR --language en
 ```
 
 * Note 1: The language options is correspond to the corpus. Currently, the tool only supports English, Simplified Chinese and Korean.
 * Note 2: Synth-Text is mainly used to generate images for OCR recognition models.
   So the height of style images should be around 32 pixels. Images in other sizes may behave poorly.
+* Note 3: You can modify `use_gpu` in `configs/config.yml` to determine whether to use GPU for prediction.
+
 
 
 For example, enter the following image and corpus `PaddleOCR`.
@@ -139,8 +141,9 @@ We provide a general dataset containing Chinese, English and Korean (50,000 imag
 2. You can run the following command to start synthesis task:
 
    ``` bash
-   python -m tools.synth_dataset.py -c configs/dataset_config.yml
+   python3 tools/synth_dataset.py -c configs/dataset_config.yml
    ```
+
 We also provide example corpus and images in `examples` folder.
     <div align="center">
         <img src="examples/style_images/1.jpg" width="300">

StyleText/README_ch.md

@@ -21,7 +21,7 @@
 </div>
 
 
-Style-Text数据合成工具是基于百度自研的文本编辑算法《Editing Text in the Wild》https://arxiv.org/abs/1908.03047
+Style-Text数据合成工具是基于百度和华科合作研发的文本编辑算法《Editing Text in the Wild》https://arxiv.org/abs/1908.03047
 
 不同于常用的基于GAN的数据合成工具，Style-Text主要框架包括：1.文本前景风格迁移模块 2.背景抽取模块 3.融合模块。经过这样三步，就可以迅速实现图像文本风格迁移。下图是一些该数据合成工具效果图。
 
@@ -61,11 +61,12 @@ fusion_generator:
 输入一张风格图和一段文字语料，运行tools/synth_image，合成单张图片，结果图像保存在当前目录下：
 
 ```python
-python3 -m tools.synth_image -c configs/config.yml --style_image examples/style_images/2.jpg --text_corpus PaddleOCR --language en
+python3 tools/synth_image.py -c configs/config.yml --style_image examples/style_images/2.jpg --text_corpus PaddleOCR --language en
 ```
 * 注1：语言选项和语料相对应，目前该工具只支持英文、简体中文和韩语。
 * 注2：Style-Text生成的数据主要应用于OCR识别场景。基于当前PaddleOCR识别模型的设计，我们主要支持高度在32左右的风格图像。
   如果输入图像尺寸相差过多，效果可能不佳。
+* 注3：可以通过修改配置文件中的`use_gpu`(true或者false)参数来决定是否使用GPU进行预测。
 
 
 例如，输入如下图片和语料"PaddleOCR":
@@ -127,7 +128,7 @@ python3 -m tools.synth_image -c configs/config.yml --style_image examples/style_
 2. 运行`tools/synth_dataset`合成数据：
 
    ``` bash
-   python -m tools.synth_dataset -c configs/dataset_config.yml
+   python3 tools/synth_dataset.py -c configs/dataset_config.yml
    ```
    我们在examples目录下提供了样例图片和语料。
     <div align="center">

StyleText/engine/predictors.py

@@ -28,6 +28,7 @@ class StyleTextRecPredictor(object):
                              ], "Generator {} not supported.".format(algorithm)
         use_gpu = config["Global"]['use_gpu']
         check_gpu(use_gpu)
+        paddle.set_device('gpu' if use_gpu else 'cpu')
         self.logger = get_logger()
         self.generator = getattr(style_text_rec, algorithm)(config)
         self.height = config["Global"]["image_height"]

StyleText/tools/synth_dataset.py

@@ -11,6 +11,14 @@
 # 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.
+
+import os
+import sys
+
+__dir__ = os.path.dirname(os.path.abspath(__file__))
+sys.path.append(__dir__)
+sys.path.append(os.path.abspath(os.path.join(__dir__, '..')))
+
 from engine.synthesisers import DatasetSynthesiser
 
 

StyleText/tools/synth_image.py

@@ -16,13 +16,13 @@ import cv2
 import sys
 import glob
 
-from utils.config import ArgsParser
-from engine.synthesisers import ImageSynthesiser
-
 __dir__ = os.path.dirname(os.path.abspath(__file__))
 sys.path.append(__dir__)
 sys.path.append(os.path.abspath(os.path.join(__dir__, '..')))
 
+from utils.config import ArgsParser
+from engine.synthesisers import ImageSynthesiser
+
 
 def synth_image():
     args = ArgsParser().parse_args()

configs/det/det_mv3_db.yml

@@ -67,7 +67,7 @@ Train:
     data_dir: ./train_data/icdar2015/text_localization/
     label_file_list:
       - ./train_data/icdar2015/text_localization/train_icdar2015_label.txt
-    ratio_list: [0.5]
+    ratio_list: [1.0]
     transforms:
       - DecodeImage: # load image
           img_mode: BGR

configs/det/det_r50_vd_db.yml

@@ -66,7 +66,7 @@ Train:
     data_dir: ./train_data/icdar2015/text_localization/
     label_file_list:
       - ./train_data/icdar2015/text_localization/train_icdar2015_label.txt
-    ratio_list: [0.5]
+    ratio_list: [1.0]
     transforms:
       - DecodeImage: # load image
           img_mode: BGR

configs/det/det_r50_vd_sast_totaltext.yml

@@ -62,7 +62,7 @@ Train:
     name: SimpleDataSet
     data_dir: ./train_data/
     label_file_list: [./train_data/art_latin_icdar_14pt/train_no_tt_test/train_label_json.txt, ./train_data/total_text_icdar_14pt/train_label_json.txt]
-    data_ratio_list: [0.5, 0.5]
+    ratio_list: [0.5, 0.5]
     transforms:
       - DecodeImage: # load image
           img_mode: BGR

deploy/cpp_infer/include/config.h

@@ -25,9 +25,9 @@
 
 namespace PaddleOCR {
 
-class Config {
+class OCRConfig {
 public:
-  explicit Config(const std::string &config_file) {
+  explicit OCRConfig(const std::string &config_file) {
     config_map_ = LoadConfig(config_file);
 
     this->use_gpu = bool(stoi(config_map_["use_gpu"]));
@@ -41,8 +41,6 @@ public:
 
     this->use_mkldnn = bool(stoi(config_map_["use_mkldnn"]));
 
-    this->use_zero_copy_run = bool(stoi(config_map_["use_zero_copy_run"]));
-
     this->max_side_len = stoi(config_map_["max_side_len"]);
 
     this->det_db_thresh = stod(config_map_["det_db_thresh"]);
@@ -76,8 +74,6 @@ public:
 
   bool use_mkldnn = false;
 
-  bool use_zero_copy_run = false;
-
   int max_side_len = 960;
 
   double det_db_thresh = 0.3;

deploy/cpp_infer/include/ocr_cls.h

@@ -30,6 +30,8 @@
 #include <include/preprocess_op.h>
 #include <include/utility.h>
 
+using namespace paddle_infer;
+
 namespace PaddleOCR {
 
 class Classifier {
@@ -37,14 +39,12 @@ public:
   explicit Classifier(const std::string &model_dir, const bool &use_gpu,
                       const int &gpu_id, const int &gpu_mem,
                       const int &cpu_math_library_num_threads,
-                      const bool &use_mkldnn, const bool &use_zero_copy_run,
-                      const double &cls_thresh) {
+                      const bool &use_mkldnn, const double &cls_thresh) {
     this->use_gpu_ = use_gpu;
     this->gpu_id_ = gpu_id;
     this->gpu_mem_ = gpu_mem;
     this->cpu_math_library_num_threads_ = cpu_math_library_num_threads;
     this->use_mkldnn_ = use_mkldnn;
-    this->use_zero_copy_run_ = use_zero_copy_run;
 
     this->cls_thresh = cls_thresh;
 
@@ -57,14 +57,13 @@ public:
   cv::Mat Run(cv::Mat &img);
 
 private:
-  std::shared_ptr<PaddlePredictor> predictor_;
+  std::shared_ptr<Predictor> predictor_;
 
   bool use_gpu_ = false;
   int gpu_id_ = 0;
   int gpu_mem_ = 4000;
   int cpu_math_library_num_threads_ = 4;
   bool use_mkldnn_ = false;
-  bool use_zero_copy_run_ = false;
   double cls_thresh = 0.5;
 
   std::vector<float> mean_ = {0.5f, 0.5f, 0.5f};

deploy/cpp_infer/include/ocr_det.h

@@ -32,6 +32,8 @@
 #include <include/postprocess_op.h>
 #include <include/preprocess_op.h>
 
+using namespace paddle_infer;
+
 namespace PaddleOCR {
 
 class DBDetector {
@@ -39,8 +41,8 @@ public:
   explicit DBDetector(const std::string &model_dir, const bool &use_gpu,
                       const int &gpu_id, const int &gpu_mem,
                       const int &cpu_math_library_num_threads,
-                      const bool &use_mkldnn, const bool &use_zero_copy_run,
-                      const int &max_side_len, const double &det_db_thresh,
+                      const bool &use_mkldnn, const int &max_side_len,
+                      const double &det_db_thresh,
                       const double &det_db_box_thresh,
                       const double &det_db_unclip_ratio,
                       const bool &visualize) {
@@ -49,7 +51,6 @@ public:
     this->gpu_mem_ = gpu_mem;
     this->cpu_math_library_num_threads_ = cpu_math_library_num_threads;
     this->use_mkldnn_ = use_mkldnn;
-    this->use_zero_copy_run_ = use_zero_copy_run;
 
     this->max_side_len_ = max_side_len;
 
@@ -69,14 +70,13 @@ public:
   void Run(cv::Mat &img, std::vector<std::vector<std::vector<int>>> &boxes);
 
 private:
-  std::shared_ptr<PaddlePredictor> predictor_;
+  std::shared_ptr<Predictor> predictor_;
 
   bool use_gpu_ = false;
   int gpu_id_ = 0;
   int gpu_mem_ = 4000;
   int cpu_math_library_num_threads_ = 4;
   bool use_mkldnn_ = false;
-  bool use_zero_copy_run_ = false;
 
   int max_side_len_ = 960;
 

deploy/cpp_infer/include/ocr_rec.h

@@ -32,6 +32,8 @@
 #include <include/preprocess_op.h>
 #include <include/utility.h>
 
+using namespace paddle_infer;
+
 namespace PaddleOCR {
 
 class CRNNRecognizer {
@@ -39,14 +41,12 @@ public:
   explicit CRNNRecognizer(const std::string &model_dir, const bool &use_gpu,
                           const int &gpu_id, const int &gpu_mem,
                           const int &cpu_math_library_num_threads,
-                          const bool &use_mkldnn, const bool &use_zero_copy_run,
-                          const string &label_path) {
+                          const bool &use_mkldnn, const string &label_path) {
     this->use_gpu_ = use_gpu;
     this->gpu_id_ = gpu_id;
     this->gpu_mem_ = gpu_mem;
     this->cpu_math_library_num_threads_ = cpu_math_library_num_threads;
     this->use_mkldnn_ = use_mkldnn;
-    this->use_zero_copy_run_ = use_zero_copy_run;
 
     this->label_list_ = Utility::ReadDict(label_path);
     this->label_list_.insert(this->label_list_.begin(),
@@ -63,14 +63,13 @@ public:
            Classifier *cls);
 
 private:
-  std::shared_ptr<PaddlePredictor> predictor_;
+  std::shared_ptr<Predictor> predictor_;
 
   bool use_gpu_ = false;
   int gpu_id_ = 0;
   int gpu_mem_ = 4000;
   int cpu_math_library_num_threads_ = 4;
   bool use_mkldnn_ = false;
-  bool use_zero_copy_run_ = false;
 
   std::vector<std::string> label_list_;
 

deploy/cpp_infer/readme.md

@@ -122,10 +122,10 @@ build/paddle_inference_install_dir/
 * 下载之后使用下面的方法解压。
 
 ```
-tar -xf fluid_inference.tgz
+tar -xf paddle_inference.tgz
 ```
 
-最终会在当前的文件夹中生成`fluid_inference/`的子文件夹。
+最终会在当前的文件夹中生成`paddle_inference/`的子文件夹。
 
 
 ## 2 开始运行
@@ -137,11 +137,11 @@ tar -xf fluid_inference.tgz
 ```
 inference/
 |-- det_db
-|   |--model
-|   |--params
+|   |--inference.pdparams
+|   |--inference.pdimodel
 |-- rec_rcnn
-|   |--model
-|   |--params
+|   |--inference.pdparams
+|   |--inference.pdparams
 ```
 
 
@@ -180,7 +180,7 @@ cmake .. \
 make -j
 ```
 
-`OPENCV_DIR`为opencv编译安装的地址；`LIB_DIR`为下载(`fluid_inference`文件夹)或者编译生成的Paddle预测库地址(`build/fluid_inference_install_dir`文件夹)；`CUDA_LIB_DIR`为cuda库文件地址，在docker中；为`/usr/local/cuda/lib64`；`CUDNN_LIB_DIR`为cudnn库文件地址，在docker中为`/usr/lib/x86_64-linux-gnu/`。
+`OPENCV_DIR`为opencv编译安装的地址；`LIB_DIR`为下载(`paddle_inference`文件夹)或者编译生成的Paddle预测库地址(`build/paddle_inference_install_dir`文件夹)；`CUDA_LIB_DIR`为cuda库文件地址，在docker中；为`/usr/local/cuda/lib64`；`CUDNN_LIB_DIR`为cudnn库文件地址，在docker中为`/usr/lib/x86_64-linux-gnu/`。
 
 
 * 编译完成之后，会在`build`文件夹下生成一个名为`ocr_system`的可执行文件。
@@ -202,7 +202,6 @@ gpu_id  0 # GPU id，使用GPU时有效
 gpu_mem  4000  # 申请的GPU内存
 cpu_math_library_num_threads  10 # CPU预测时的线程数，在机器核数充足的情况下，该值越大，预测速度越快
 use_mkldnn 1 # 是否使用mkldnn库
-use_zero_copy_run 1 # 是否使用use_zero_copy_run进行预测
 
 # det config
 max_side_len  960 # 输入图像长宽大于960时，等比例缩放图像，使得图像最长边为960

deploy/cpp_infer/readme_en.md

@@ -107,10 +107,10 @@ make inference_lib_dist
 For more compilation parameter options, please refer to the official website of the Paddle C++ inference library:[https://www.paddlepaddle.org.cn/documentation/docs/en/advanced_guide/inference_deployment/inference/build_and_install_lib_en.html](https://www.paddlepaddle.org.cn/documentation/docs/en/advanced_guide/inference_deployment/inference/build_and_install_lib_en.html).
 
 
-* After the compilation process, you can see the following files in the folder of `build/fluid_inference_install_dir/`.
+* After the compilation process, you can see the following files in the folder of `build/paddle_inference_install_dir/`.
 
 ```
-build/fluid_inference_install_dir/
+build/paddle_inference_install_dir/
 |-- CMakeCache.txt
 |-- paddle
 |-- third_party
@@ -130,10 +130,10 @@ Among them, `paddle` is the Paddle library required for C++ prediction later, an
 * After downloading, use the following method to uncompress.
 
 ```
-tar -xf fluid_inference.tgz
+tar -xf paddle_inference.tgz
 ```
 
-Finally you can see the following files in the folder of `fluid_inference/`.
+Finally you can see the following files in the folder of `paddle_inference/`.
 
 
 ## 2. Compile and run the demo
@@ -145,11 +145,11 @@ Finally you can see the following files in the folder of `fluid_inference/`.
 ```
 inference/
 |-- det_db
-|   |--model
-|   |--params
+|   |--inference.pdparams
+|   |--inference.pdimodel
 |-- rec_rcnn
-|   |--model
-|   |--params
+|   |--inference.pdparams
+|   |--inference.pdparams
 ```
 
 
@@ -188,7 +188,9 @@ cmake .. \
 make -j
 ```
 
-`OPENCV_DIR` is the opencv installation path; `LIB_DIR` is the download (`fluid_inference` folder) or the generated Paddle inference library path (`build/fluid_inference_install_dir` folder); `CUDA_LIB_DIR` is the cuda library file path, in docker; it is `/usr/local/cuda/lib64`; `CUDNN_LIB_DIR` is the cudnn library file path, in docker it is `/usr/lib/x86_64-linux-gnu/`.
+`OPENCV_DIR` is the opencv installation path; `LIB_DIR` is the download (`paddle_inference` folder)
+or the generated Paddle inference library path (`build/paddle_inference_install_dir` folder);
+`CUDA_LIB_DIR` is the cuda library file path, in docker; it is `/usr/local/cuda/lib64`; `CUDNN_LIB_DIR` is the cudnn library file path, in docker it is `/usr/lib/x86_64-linux-gnu/`.
 
 
 * After the compilation is completed, an executable file named `ocr_system` will be generated in the `build` folder.
@@ -211,7 +213,6 @@ gpu_id  0 # GPU id when use_gpu is 1
 gpu_mem  4000  # GPU memory requested
 cpu_math_library_num_threads  10 # Number of threads when using CPU inference. When machine cores is enough, the large the value, the faster the inference speed
 use_mkldnn 1 # Whether to use mkdlnn library
-use_zero_copy_run 1 # Whether to use use_zero_copy_run for inference
 
 max_side_len  960 #  Limit the maximum image height and width to 960
 det_db_thresh  0.3 # Used to filter the binarized image of DB prediction, setting 0.-0.3 has no obvious effect on the result
@@ -244,4 +245,4 @@ The detection results will be shown on the screen, which is as follows.
 
 ### 2.3 Notes
 
-* Paddle2.0.0-beta0 inference model library is recommanded for this tuturial.
+* Paddle2.0.0-beta0 inference model library is recommended for this toturial.

deploy/cpp_infer/src/config.cpp

@@ -16,7 +16,7 @@
 
 namespace PaddleOCR {
 
-std::vector<std::string> Config::split(const std::string &str,
+std::vector<std::string> OCRConfig::split(const std::string &str,
                                           const std::string &delim) {
   std::vector<std::string> res;
   if ("" == str)
@@ -38,7 +38,7 @@ std::vector<std::string> Config::split(const std::string &str,
 }
 
 std::map<std::string, std::string>
-Config::LoadConfig(const std::string &config_path) {
+OCRConfig::LoadConfig(const std::string &config_path) {
   auto config = Utility::ReadDict(config_path);
 
   std::map<std::string, std::string> dict;
@@ -53,7 +53,7 @@ Config::LoadConfig(const std::string &config_path) {
   return dict;
 }
 
-void Config::PrintConfigInfo() {
+void OCRConfig::PrintConfigInfo() {
   std::cout << "=======Paddle OCR inference config======" << std::endl;
   for (auto iter = config_map_.begin(); iter != config_map_.end(); iter++) {
     std::cout << iter->first << " : " << iter->second << std::endl;

deploy/cpp_infer/src/main.cpp

@@ -42,7 +42,7 @@ int main(int argc, char **argv) {
     exit(1);
   }
 
-  Config config(argv[1]);
+  OCRConfig config(argv[1]);
 
   config.PrintConfigInfo();
 
@@ -50,37 +50,22 @@ int main(int argc, char **argv) {
 
   cv::Mat srcimg = cv::imread(img_path, cv::IMREAD_COLOR);
 
-  DBDetector det(
-      config.det_model_dir, config.use_gpu, config.gpu_id, config.gpu_mem,
-      config.cpu_math_library_num_threads, config.use_mkldnn,
-      config.use_zero_copy_run, config.max_side_len, config.det_db_thresh,
-      config.det_db_box_thresh, config.det_db_unclip_ratio, config.visualize);
+  DBDetector det(config.det_model_dir, config.use_gpu, config.gpu_id,
+                 config.gpu_mem, config.cpu_math_library_num_threads,
+                 config.use_mkldnn, config.max_side_len, config.det_db_thresh,
+                 config.det_db_box_thresh, config.det_db_unclip_ratio,
+                 config.visualize);
 
   Classifier *cls = nullptr;
   if (config.use_angle_cls == true) {
     cls = new Classifier(config.cls_model_dir, config.use_gpu, config.gpu_id,
                          config.gpu_mem, config.cpu_math_library_num_threads,
-                         config.use_mkldnn, config.use_zero_copy_run,
-                         config.cls_thresh);
+                         config.use_mkldnn, config.cls_thresh);
   }
 
   CRNNRecognizer rec(config.rec_model_dir, config.use_gpu, config.gpu_id,
                      config.gpu_mem, config.cpu_math_library_num_threads,
-                     config.use_mkldnn, config.use_zero_copy_run,
-                     config.char_list_file);
-
-#ifdef USE_MKL
-#pragma omp parallel
-  for (auto i = 0; i < 10; i++) {
-    LOG_IF(WARNING,
-           config.cpu_math_library_num_threads != omp_get_num_threads())
-        << "WARNING! MKL is running on " << omp_get_num_threads()
-        << " threads while cpu_math_library_num_threads is set to "
-        << config.cpu_math_library_num_threads
-        << ". Possible reason could be 1. You have set omp_set_num_threads() "
-           "somewhere; 2. MKL is not linked properly";
-  }
-#endif
+                     config.use_mkldnn, config.char_list_file);
 
   auto start = std::chrono::system_clock::now();
   std::vector<std::vector<std::vector<int>>> boxes;

deploy/cpp_infer/src/ocr_cls.cpp

@@ -35,26 +35,16 @@ cv::Mat Classifier::Run(cv::Mat &img) {
   this->permute_op_.Run(&resize_img, input.data());
 
   // Inference.
-  if (this->use_zero_copy_run_) {
   auto input_names = this->predictor_->GetInputNames();
-    auto input_t = this->predictor_->GetInputTensor(input_names[0]);
+  auto input_t = this->predictor_->GetInputHandle(input_names[0]);
   input_t->Reshape({1, 3, resize_img.rows, resize_img.cols});
-    input_t->copy_from_cpu(input.data());
-    this->predictor_->ZeroCopyRun();
-  } else {
-    paddle::PaddleTensor input_t;
-    input_t.shape = {1, 3, resize_img.rows, resize_img.cols};
-    input_t.data =
-        paddle::PaddleBuf(input.data(), input.size() * sizeof(float));
-    input_t.dtype = PaddleDType::FLOAT32;
-    std::vector<paddle::PaddleTensor> outputs;
-    this->predictor_->Run({input_t}, &outputs, 1);
-  }
+  input_t->CopyFromCpu(input.data());
+  this->predictor_->Run();
 
   std::vector<float> softmax_out;
   std::vector<int64_t> label_out;
   auto output_names = this->predictor_->GetOutputNames();
-  auto softmax_out_t = this->predictor_->GetOutputTensor(output_names[0]);
+  auto softmax_out_t = this->predictor_->GetOutputHandle(output_names[0]);
   auto softmax_shape_out = softmax_out_t->shape();
 
   int softmax_out_num =
@@ -63,7 +53,7 @@ cv::Mat Classifier::Run(cv::Mat &img) {
 
   softmax_out.resize(softmax_out_num);
 
-  softmax_out_t->copy_to_cpu(softmax_out.data());
+  softmax_out_t->CopyToCpu(softmax_out.data());
 
   float score = 0;
   int label = 0;
@@ -95,7 +85,7 @@ void Classifier::LoadModel(const std::string &model_dir) {
   }
 
   // false for zero copy tensor
-  config.SwitchUseFeedFetchOps(!this->use_zero_copy_run_);
+  config.SwitchUseFeedFetchOps(false);
   // true for multiple input
   config.SwitchSpecifyInputNames(true);
 
@@ -104,6 +94,6 @@ void Classifier::LoadModel(const std::string &model_dir) {
   config.EnableMemoryOptim();
   config.DisableGlogInfo();
 
-  this->predictor_ = CreatePaddlePredictor(config);
+  this->predictor_ = CreatePredictor(config);
 }
 } // namespace PaddleOCR

deploy/cpp_infer/src/ocr_det.cpp

@@ -17,12 +17,17 @@
 namespace PaddleOCR {
 
 void DBDetector::LoadModel(const std::string &model_dir) {
-  AnalysisConfig config;
+  //   AnalysisConfig config;
+  paddle_infer::Config config;
   config.SetModel(model_dir + "/inference.pdmodel",
                   model_dir + "/inference.pdiparams");
 
   if (this->use_gpu_) {
     config.EnableUseGpu(this->gpu_mem_, this->gpu_id_);
+    //     config.EnableTensorRtEngine(
+    //           1 << 20, 1, 3,
+    //           AnalysisConfig::Precision::kFloat32,
+    //           false, false);
   } else {
     config.DisableGpu();
     if (this->use_mkldnn_) {
@@ -32,10 +37,8 @@ void DBDetector::LoadModel(const std::string &model_dir) {
     }
     config.SetCpuMathLibraryNumThreads(this->cpu_math_library_num_threads_);
   }
-
-  // false for zero copy tensor
-  // true for commom tensor
-  config.SwitchUseFeedFetchOps(!this->use_zero_copy_run_);
+  // use zero_copy_run as default
+  config.SwitchUseFeedFetchOps(false);
   // true for multiple input
   config.SwitchSpecifyInputNames(true);
 
@@ -44,7 +47,7 @@ void DBDetector::LoadModel(const std::string &model_dir) {
   config.EnableMemoryOptim();
   config.DisableGlogInfo();
 
-  this->predictor_ = CreatePaddlePredictor(config);
+  this->predictor_ = CreatePredictor(config);
 }
 
 void DBDetector::Run(cv::Mat &img,
@@ -64,31 +67,21 @@ void DBDetector::Run(cv::Mat &img,
   this->permute_op_.Run(&resize_img, input.data());
 
   // Inference.
-  if (this->use_zero_copy_run_) {
   auto input_names = this->predictor_->GetInputNames();
-    auto input_t = this->predictor_->GetInputTensor(input_names[0]);
+  auto input_t = this->predictor_->GetInputHandle(input_names[0]);
   input_t->Reshape({1, 3, resize_img.rows, resize_img.cols});
-    input_t->copy_from_cpu(input.data());
-    this->predictor_->ZeroCopyRun();
-  } else {
-    paddle::PaddleTensor input_t;
-    input_t.shape = {1, 3, resize_img.rows, resize_img.cols};
-    input_t.data =
-        paddle::PaddleBuf(input.data(), input.size() * sizeof(float));
-    input_t.dtype = PaddleDType::FLOAT32;
-    std::vector<paddle::PaddleTensor> outputs;
-    this->predictor_->Run({input_t}, &outputs, 1);
-  }
+  input_t->CopyFromCpu(input.data());
+  this->predictor_->Run();
 
   std::vector<float> out_data;
   auto output_names = this->predictor_->GetOutputNames();
-  auto output_t = this->predictor_->GetOutputTensor(output_names[0]);
+  auto output_t = this->predictor_->GetOutputHandle(output_names[0]);
   std::vector<int> output_shape = output_t->shape();
   int out_num = std::accumulate(output_shape.begin(), output_shape.end(), 1,
                                 std::multiplies<int>());
 
   out_data.resize(out_num);
-  output_t->copy_to_cpu(out_data.data());
+  output_t->CopyToCpu(out_data.data());
 
   int n2 = output_shape[2];
   int n3 = output_shape[3];