Merge remote-tracking branch 'PaddlePaddle/dygraph' into dygraph

2025-06-03 21:53:39 +08:00 · 2021-05-04 23:18:56 +08:00 · 2021-05-04 23:18:56 +08:00 · 56feca0106
commit 56feca0106
parent bfea4a63af c4d20782cf
20 changed files with 211 additions and 320 deletions
--- a/README_ch.md
+++ b/README_ch.md
@ -8,7 +8,6 @@ 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**
- 【预告】 PaddleOCR研发团队对最新发版内容技术深入解读，4月13日晚上19:00，[直播地址](https://live.bilibili.com/21689802)
 - 2021.4.8 release 2.1版本，新增AAAI 2021论文[端到端识别算法PGNet](./doc/doc_ch/pgnet.md)开源，[多语言模型](./doc/doc_ch/multi_languages.md)支持种类增加到80+。
 - 2021.2.1 [FAQ](./doc/doc_ch/FAQ.md)新增5个高频问题，总数162个，每周一都会更新，欢迎大家持续关注。
 - 2021.1.21 更新多语言识别模型，目前支持语种超过27种，包括中文简体、中文繁体、英文、法文、德文、韩文、日文、意大利文、西班牙文、葡萄牙文、俄罗斯文、阿拉伯文等，后续计划可以参考[多语言研发计划](https://github.com/PaddlePaddle/PaddleOCR/issues/1048)
@ -80,7 +79,7 @@ PaddleOCR同时支持动态图与静态图两种编程范式
 - 算法介绍
    - [文本检测](./doc/doc_ch/algorithm_overview.md)
    - [文本识别](./doc/doc_ch/algorithm_overview.md)
-    - [PP-OCR Pipline](#PP-OCR)
+    - [PP-OCR Pipeline](#PP-OCR)
    - [端到端PGNet算法](./doc/doc_ch/pgnet.md)
 - 模型训练/评估
    - [文本检测](./doc/doc_ch/detection.md)
@ -115,7 +114,7 @@ PaddleOCR同时支持动态图与静态图两种编程范式


 <a name="PP-OCR"></a>
-## PP-OCR Pipline
+## PP-OCR Pipeline
 <div align="center">
    <img src="./doc/ppocr_framework.png" width="800">
 </div>
--- a/deploy/cpp_infer/include/config.h
+++ b/deploy/cpp_infer/include/config.h
@ -49,6 +49,8 @@ public:

    this->det_db_unclip_ratio = stod(config_map_["det_db_unclip_ratio"]);

+    this->use_polygon_score = bool(stoi(config_map_["use_polygon_score"]));
+
    this->det_model_dir.assign(config_map_["det_model_dir"]);

    this->rec_model_dir.assign(config_map_["rec_model_dir"]);
@ -86,6 +88,8 @@ public:

  double det_db_unclip_ratio = 2.0;

+  bool use_polygon_score = false;
+
  std::string det_model_dir;

  std::string rec_model_dir;
--- a/deploy/cpp_infer/include/ocr_det.h
+++ b/deploy/cpp_infer/include/ocr_det.h
@ -44,7 +44,8 @@ public:
                      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,
+                      const double &det_db_unclip_ratio,
+                      const bool &use_polygon_score, const bool &visualize,
                      const bool &use_tensorrt, const bool &use_fp16) {
    this->use_gpu_ = use_gpu;
    this->gpu_id_ = gpu_id;
@ -57,6 +58,7 @@ public:
    this->det_db_thresh_ = det_db_thresh;
    this->det_db_box_thresh_ = det_db_box_thresh;
    this->det_db_unclip_ratio_ = det_db_unclip_ratio;
+    this->use_polygon_score_ = use_polygon_score;

    this->visualize_ = visualize;
    this->use_tensorrt_ = use_tensorrt;
@ -85,6 +87,7 @@ private:
  double det_db_thresh_ = 0.3;
  double det_db_box_thresh_ = 0.5;
  double det_db_unclip_ratio_ = 2.0;
+  bool use_polygon_score_ = false;

  bool visualize_ = true;
  bool use_tensorrt_ = false;
--- a/deploy/cpp_infer/include/postprocess_op.h
+++ b/deploy/cpp_infer/include/postprocess_op.h
@ -55,7 +55,8 @@ public:

  std::vector<std::vector<std::vector<int>>>
  BoxesFromBitmap(const cv::Mat pred, const cv::Mat bitmap,
-                  const float &box_thresh, const float &det_db_unclip_ratio);
+                  const float &box_thresh, const float &det_db_unclip_ratio,
+                  const bool &use_polygon_score);

  std::vector<std::vector<std::vector<int>>>
  FilterTagDetRes(std::vector<std::vector<std::vector<int>>> boxes,
--- a/deploy/cpp_infer/readme.md
+++ b/deploy/cpp_infer/readme.md
@ -183,7 +183,7 @@ cmake .. \
 make -j
 ```

-`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/`。
+`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。
@ -211,6 +211,7 @@ max_side_len  960 # 输入图像长宽大于960时，等比例缩放图像，使
 det_db_thresh  0.3 # 用于过滤DB预测的二值化图像，设置为0.-0.3对结果影响不明显
 det_db_box_thresh  0.5 # DB后处理过滤box的阈值，如果检测存在漏框情况，可酌情减小
 det_db_unclip_ratio  1.6 # 表示文本框的紧致程度，越小则文本框更靠近文本
+use_polygon_score 1 # 是否使用多边形框计算bbox score，0表示使用矩形框计算。矩形框计算速度更快，多边形框对弯曲文本区域计算更准确。
 det_model_dir  ./inference/det_db # 检测模型inference model地址

 # cls config
--- a/deploy/cpp_infer/readme_en.md
+++ b/deploy/cpp_infer/readme_en.md
@ -217,6 +217,7 @@ 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
 det_db_box_thresh  0.5 # DDB post-processing filter box threshold, if there is a missing box detected, it can be reduced as appropriate
 det_db_unclip_ratio  1.6 # Indicates the compactness of the text box, the smaller the value, the closer the text box to the text
+use_polygon_score 1 # Whether to use polygon box to calculate bbox score, 0 means to use rectangle box to calculate. Use rectangular box to calculate faster, and polygonal box more accurate for curved text area.
 det_model_dir  ./inference/det_db # Address of detection inference model

 # cls config
--- a/deploy/cpp_infer/src/main.cpp
+++ b/deploy/cpp_infer/src/main.cpp
@ -59,7 +59,8 @@ int main(int argc, char **argv) {
                 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, config.use_tensorrt, config.use_fp16);
+                 config.use_polygon_score, config.visualize,
+                 config.use_tensorrt, config.use_fp16);

  Classifier *cls = nullptr;
  if (config.use_angle_cls == true) {
--- a/deploy/cpp_infer/src/ocr_det.cpp
+++ b/deploy/cpp_infer/src/ocr_det.cpp
@ -109,9 +109,9 @@ void DBDetector::Run(cv::Mat &img,
  cv::Mat dilation_map;
  cv::Mat dila_ele = cv::getStructuringElement(cv::MORPH_RECT, cv::Size(2, 2));
  cv::dilate(bit_map, dilation_map, dila_ele);
-  boxes = post_processor_.BoxesFromBitmap(pred_map, dilation_map,
-                                          this->det_db_box_thresh_,
-                                          this->det_db_unclip_ratio_);
+  boxes = post_processor_.BoxesFromBitmap(
+      pred_map, dilation_map, this->det_db_box_thresh_,
+      this->det_db_unclip_ratio_, this->use_polygon_score_);

  boxes = post_processor_.FilterTagDetRes(boxes, ratio_h, ratio_w, srcimg);

--- a/deploy/cpp_infer/src/postprocess_op.cpp
+++ b/deploy/cpp_infer/src/postprocess_op.cpp
@ -160,35 +160,49 @@ std::vector<std::vector<float>> PostProcessor::GetMiniBoxes(cv::RotatedRect box,
 }

 float PostProcessor::PolygonScoreAcc(std::vector<cv::Point> contour,
-		                  cv::Mat pred){
+                                     cv::Mat pred) {
  int width = pred.cols;
  int height = pred.rows;
  std::vector<float> box_x;
  std::vector<float> box_y;
-  for(int i=0; i<contour.size(); ++i){
+  for (int i = 0; i < contour.size(); ++i) {
    box_x.push_back(contour[i].x);
    box_y.push_back(contour[i].y);
  }

-  int xmin = clamp(int(std::floor(*(std::min_element(box_x.begin(), box_x.end())))), 0, width - 1);
-  int xmax = clamp(int(std::ceil(*(std::max_element(box_x.begin(), box_x.end())))), 0, width - 1);
-  int ymin = clamp(int(std::floor(*(std::min_element(box_y.begin(), box_y.end())))), 0, height - 1);
-  int ymax = clamp(int(std::ceil(*(std::max_element(box_y.begin(), box_y.end())))), 0, height - 1);
+  int xmin =
+      clamp(int(std::floor(*(std::min_element(box_x.begin(), box_x.end())))), 0,
+            width - 1);
+  int xmax =
+      clamp(int(std::ceil(*(std::max_element(box_x.begin(), box_x.end())))), 0,
+            width - 1);
+  int ymin =
+      clamp(int(std::floor(*(std::min_element(box_y.begin(), box_y.end())))), 0,
+            height - 1);
+  int ymax =
+      clamp(int(std::ceil(*(std::max_element(box_y.begin(), box_y.end())))), 0,
+            height - 1);

  cv::Mat mask;
  mask = cv::Mat::zeros(ymax - ymin + 1, xmax - xmin + 1, CV_8UC1);

-  cv::Point rook_point[contour.size()];
-  for(int i=0; i<contour.size(); ++i){
+
+  cv::Point* rook_point = new cv::Point[contour.size()];
+   
+  for (int i = 0; i < contour.size(); ++i) {
    rook_point[i] = cv::Point(int(box_x[i]) - xmin, int(box_y[i]) - ymin);
  }
  const cv::Point *ppt[1] = {rook_point};
  int npt[] = {int(contour.size())};
+
+
  cv::fillPoly(mask, ppt, npt, 1, cv::Scalar(1));

  cv::Mat croppedImg;
  pred(cv::Rect(xmin, ymin, xmax - xmin + 1, ymax - ymin + 1)).copyTo(croppedImg);
  float score = cv::mean(croppedImg, mask)[0];
+
+  delete []rook_point;
  return score;
 }

@ -230,10 +244,9 @@ float PostProcessor::BoxScoreFast(std::vector<std::vector<float>> box_array,
  return score;
 }

-std::vector<std::vector<std::vector<int>>>
-PostProcessor::BoxesFromBitmap(const cv::Mat pred, const cv::Mat bitmap,
-                               const float &box_thresh,
-                               const float &det_db_unclip_ratio) {
+std::vector<std::vector<std::vector<int>>> PostProcessor::BoxesFromBitmap(
+    const cv::Mat pred, const cv::Mat bitmap, const float &box_thresh,
+    const float &det_db_unclip_ratio, const bool &use_polygon_score) {
  const int min_size = 3;
  const int max_candidates = 1000;

@ -267,9 +280,12 @@ PostProcessor::BoxesFromBitmap(const cv::Mat pred, const cv::Mat bitmap,
    }

    float score;
-    score = BoxScoreFast(array, pred);
-    /* compute using polygon*/ 
-    // score = PolygonScoreAcc(contours[_i], pred);
+    if (use_polygon_score)
+      /* compute using polygon*/
+      score = PolygonScoreAcc(contours[_i], pred);
+    else
+      score = BoxScoreFast(array, pred);
+
    if (score < box_thresh)
      continue;

--- a/deploy/cpp_infer/tools/config.txt
+++ b/deploy/cpp_infer/tools/config.txt
@ -10,6 +10,7 @@ max_side_len  960
 det_db_thresh  0.3
 det_db_box_thresh  0.5
 det_db_unclip_ratio  1.6
+use_polygon_score 1
 det_model_dir  ./inference/ch_ppocr_mobile_v2.0_det_infer/

 # cls config
--- a/doc/doc_ch/multi_languages.md
+++ b/doc/doc_ch/multi_languages.md
@ -11,7 +11,7 @@ PaddleOCR 旨在打造一套丰富、领先、且实用的OCR工具库，不仅
 其中英文模型支持，大小写字母和常见标点的检测识别，并优化了空格字符的识别：

 <div align="center">
-    <img src="../imgs_results/multi_lang/en_1.jpg" width="400" height="600">
+    <img src="../imgs_results/multi_lang/img_12.jpg" width="900" height="300">
 </div>

 小语种模型覆盖了拉丁语系、阿拉伯语系、中文繁体、韩语、日语等等：
@ -19,6 +19,8 @@ PaddleOCR 旨在打造一套丰富、领先、且实用的OCR工具库，不仅
 <div align="center">
    <img src="../imgs_results/multi_lang/japan_2.jpg" width="600" height="300">
    <img src="../imgs_results/multi_lang/french_0.jpg" width="300" height="300">
+    <img src="../imgs_results/multi_lang/korean_0.jpg" width="500" height="300">
+    <img src="../imgs_results/multi_lang/arabic_0.jpg" width="300" height="300">
 </div>


@ -30,14 +32,9 @@ PaddleOCR 旨在打造一套丰富、领先、且实用的OCR工具库，不仅

 - [2 快速使用](#快速使用)
    - [2.1 命令行运行](#命令行运行)
-     - [2.1.1 整图预测](#bash_检测+识别)
-     - [2.1.2 识别预测](#bash_识别)
-     - [2.1.3 检测预测](#bash_检测)
    - [2.2 python 脚本运行](#python_脚本运行)
-     - [2.2.1 整图预测](#python_检测+识别)
-     - [2.2.2 识别预测](#python_识别)
-     - [2.2.3 检测预测](#python_检测)
 - [3 自定义训练](#自定义训练)
+- [4 预测部署](#预测部署)
 - [4 支持语种及缩写](#语种缩写)

 <a name="安装"></a>
@ -50,7 +47,7 @@ PaddleOCR 旨在打造一套丰富、领先、且实用的OCR工具库，不仅
 pip install paddlepaddle

 # gpu
-pip instll paddlepaddle-gpu
+pip install paddlepaddle-gpu
 ```

 <a name="paddleocr_package_安装"></a>
@ -108,8 +105,6 @@ paddleocr --image_dir doc/imgs/japan_2.jpg --lang=japan
 paddleocr --image_dir doc/imgs_words/japan/1.jpg   --det false --lang=japan
 ```

-![](https://raw.githubusercontent.com/PaddlePaddle/PaddleOCR/release/2.0/doc/imgs_words/japan/1.jpg)
-
 结果是一个tuple，返回识别结果和识别置信度

 ```text
@ -145,6 +140,9 @@ from paddleocr import PaddleOCR, draw_ocr
 ocr = PaddleOCR(lang="korean") # 首次执行会自动下载模型文件
 img_path = 'doc/imgs/korean_1.jpg '
 result = ocr.ocr(img_path)
+# 可通过参数控制单独执行识别、检测
+# result = ocr.ocr(img_path, det=False) 只执行识别
+# result = ocr.ocr(img_path, rec=False) 只执行检测
 # 打印检测框和识别结果
 for line in result:
    print(line)
@ -166,59 +164,7 @@ im_show.save('result.jpg')
    <img src="https://raw.githubusercontent.com/PaddlePaddle/PaddleOCR/release/2.1/doc/imgs_results/korean.jpg" width="800">
 </div>

-* 识别预测
-
-```
-from paddleocr import PaddleOCR
-ocr = PaddleOCR(lang="german")
-img_path = 'PaddleOCR/doc/imgs_words/german/1.jpg'
-result = ocr.ocr(img_path, det=False, cls=True)
-for line in result:
-    print(line)
-```
-
-
-![](../imgs_words/german/1.jpg)
-
-结果是一个tuple，只包含识别结果和识别置信度
-
-```
-('leider auch jetzt', 0.97538936)
-```
-
-* 检测预测
-
-```python
-from paddleocr import PaddleOCR, draw_ocr
-ocr = PaddleOCR() # need to run only once to download and load model into memory
-img_path = 'PaddleOCR/doc/imgs_en/img_12.jpg'
-result = ocr.ocr(img_path, rec=False)
-for line in result:
-    print(line)
-
-# 显示结果
-from PIL import Image
-
-image = Image.open(img_path).convert('RGB')
-im_show = draw_ocr(image, result, txts=None, scores=None, font_path='/path/to/PaddleOCR/doc/fonts/simfang.ttf')
-im_show = Image.fromarray(im_show)
-im_show.save('result.jpg')
-```
-结果是一个list，每个item只包含文本框
-```bash
-[[26.0, 457.0], [137.0, 457.0], [137.0, 477.0], [26.0, 477.0]]
-[[25.0, 425.0], [372.0, 425.0], [372.0, 448.0], [25.0, 448.0]]
-[[128.0, 397.0], [273.0, 397.0], [273.0, 414.0], [128.0, 414.0]]
-......
-```
-
-结果可视化 ：
-
-<div align="center">
-    <img src="https://raw.githubusercontent.com/PaddlePaddle/PaddleOCR/release/2.1/doc/imgs_results/whl/12_det.jpg" width="800">
-</div>
-
-ppocr 还支持方向分类， 更多使用方式请参考：[whl包使用说明](https://github.com/PaddlePaddle/PaddleOCR/blob/release/2.0/doc/doc_ch/whl.md)。
+ppocr 还支持方向分类， 更多使用方式请参考：[whl包使用说明](https://github.com/PaddlePaddle/PaddleOCR/blob/release/2.0/doc/doc_ch/whl.md)

 <a name="自定义训练"></a>
 ## 3 自定义训练
@ -229,84 +175,58 @@ ppocr 支持使用自己的数据进行自定义训练或finetune, 其中识别
 具体数据准备、训练过程可参考：[文本检测](../doc_ch/detection.md)、[文本识别](../doc_ch/recognition.md)，更多功能如预测部署、
 数据标注等功能可以阅读完整的[文档教程](../../README_ch.md)。

-<a name="语种缩写"></a>
-## 4 支持语种及缩写
+<a name="预测部署"></a>
+## 4 预测部署

-| 语种 | 描述 | 缩写 |
-| --- | --- | --- |
-|中文|chinese and english|ch|
-|英文|english|en|
-|法文|french|fr|
-|德文|german|german|
-|日文|japan|japan|
-|韩文|korean|korean|
-|中文繁体|chinese traditional |chinese_cht|
-|意大利文| Italian |it|
-|西班牙文|Spanish |es|
-|葡萄牙文| Portuguese|pt|
-|俄罗斯文|Russia|ru|
-|阿拉伯文|Arabic|ar|
-|印地文|Hindi|hi|
-|维吾尔|Uyghur|ug|
-|波斯文|Persian|fa|
-|乌尔都文|Urdu|ur|
-|塞尔维亚文（latin)| Serbian(latin) |rs_latin|
-|欧西坦文|Occitan |oc|
-|马拉地文|Marathi|mr|
-|尼泊尔文|Nepali|ne|
-|塞尔维亚文（cyrillic)|Serbian(cyrillic)|rs_cyrillic|
-|保加利亚文|Bulgarian |bg|
-|乌克兰文|Ukranian|uk|
-|白俄罗斯文|Belarusian|be|
-|泰卢固文|Telugu |te|
-|泰米尔文|Tamil |ta|
-|南非荷兰文 |Afrikaans |af|
-|阿塞拜疆文 |Azerbaijani  |az|
-|波斯尼亚文|Bosnian|bs|
-|捷克文|Czech|cs|
-|威尔士文 |Welsh |cy|
-|丹麦文 |Danish|da|
-|爱沙尼亚文 |Estonian |et|
-|爱尔兰文 |Irish |ga|
-|克罗地亚文|Croatian |hr|
-|匈牙利文|Hungarian |hu|
-|印尼文|Indonesian|id|
-|冰岛文 |Icelandic|is|
-|库尔德文 |Kurdish|ku|
-|立陶宛文|Lithuanian |lt|
-|拉脱维亚文 |Latvian |lv|
-|毛利文|Maori|mi|
-|马来文 |Malay|ms|
-|马耳他文 |Maltese |mt|
-|荷兰文 |Dutch |nl|
-|挪威文 |Norwegian |no|
-|波兰文|Polish |pl|
-| 罗马尼亚文|Romanian |ro|
-| 斯洛伐克文|Slovak |sk|
-| 斯洛文尼亚文|Slovenian |sl|
-| 阿尔巴尼亚文|Albanian |sq|
-| 瑞典文|Swedish |sv|
-| 西瓦希里文|Swahili |sw|
-| 塔加洛文|Tagalog |tl|
-| 土耳其文|Turkish |tr|
-| 乌兹别克文|Uzbek |uz|
-| 越南文|Vietnamese |vi|
-| 蒙古文|Mongolian |mn|
-| 阿巴扎文|Abaza |abq|
-| 阿迪赫文|Adyghe |ady|
-| 卡巴丹文|Kabardian |kbd|
-| 阿瓦尔文|Avar |ava|
-| 达尔瓦文|Dargwa |dar|
-| 因古什文|Ingush |inh|
-| 拉克文|Lak |lbe|
-| 莱兹甘文|Lezghian |lez|
-|塔巴萨兰文 |Tabassaran |tab|
-| 比尔哈文|Bihari |bh|
-| 迈蒂利文|Maithili |mai|
-| 昂加文|Angika |ang|
-| 孟加拉文|Bhojpuri |bho|
-| 摩揭陀文 |Magahi |mah|
-| 那格浦尔文|Nagpur |sck|
-| 尼瓦尔文|Newari |new|
-| 保加利亚文 |Goan Konkani|gom|
-| 沙特阿拉伯文|Saudi Arabia|sa|
+除了安装whl包进行快速预测，ppocr 也提供了多种预测部署方式，如有需求可阅读相关文档：
+- [基于Python脚本预测引擎推理](./inference.md)
+- [基于C++预测引擎推理](../../deploy/cpp_infer/readme.md)
+- [服务化部署](../../deploy/hubserving/readme.md)
+- [端侧部署](https://github.com/PaddlePaddle/PaddleOCR/blob/develop/deploy/lite/readme.md)
+- [Benchmark](./benchmark.md)
+
+
+
+<a name="语种缩写"></a>
+## 5 支持语种及缩写
+
+| 语种 | 描述 | 缩写 | | 语种 | 描述 | 缩写 |
+| --- | --- | --- | ---|--- | --- | --- |
+|中文|chinese and english|ch| |保加利亚文|Bulgarian |bg|
+|英文|english|en| |乌克兰文|Ukranian|uk|
+|法文|french|fr| |白俄罗斯文|Belarusian|be|
+|德文|german|german| |泰卢固文|Telugu |te|
+|日文|japan|japan| | |阿巴扎文|Abaza |abq|
+|韩文|korean|korean| |泰米尔文|Tamil |ta|
+|中文繁体|chinese traditional |ch_tra| |南非荷兰文 |Afrikaans |af|
+|意大利文| Italian |it| |阿塞拜疆文 |Azerbaijani    |az|
+|西班牙文|Spanish |es| |波斯尼亚文|Bosnian|bs|
+|葡萄牙文| Portuguese|pt| |捷克文|Czech|cs|
+|俄罗斯文|Russia|ru| |威尔士文 |Welsh |cy|
+|阿拉伯文|Arabic|ar| |丹麦文 |Danish|da|
+|印地文|Hindi|hi| |爱沙尼亚文 |Estonian |et|
+|维吾尔|Uyghur|ug| |爱尔兰文 |Irish |ga|
+|波斯文|Persian|fa| |克罗地亚文|Croatian |hr|
+|乌尔都文|Urdu|ur| |匈牙利文|Hungarian |hu|
+|塞尔维亚文（latin)| Serbian(latin) |rs_latin| |印尼文|Indonesian|id|
+|欧西坦文|Occitan |oc| |冰岛文 |Icelandic|is|
+|马拉地文|Marathi|mr| |库尔德文 |Kurdish|ku|
+|尼泊尔文|Nepali|ne| |立陶宛文|Lithuanian |lt|
+|塞尔维亚文（cyrillic)|Serbian(cyrillic)|rs_cyrillic| |拉脱维亚文 |Latvian |lv|
+|毛利文|Maori|mi| | 达尔瓦文|Dargwa |dar|
+|马来文 |Malay|ms| | 因古什文|Ingush |inh|
+|马耳他文 |Maltese |mt| | 拉克文|Lak |lbe|
+|荷兰文 |Dutch |nl| | 莱兹甘文|Lezghian |lez|
+|挪威文 |Norwegian |no| |塔巴萨兰文 |Tabassaran |tab|
+|波兰文|Polish |pl| | 比尔哈文|Bihari |bh|
+| 罗马尼亚文|Romanian |ro| | 迈蒂利文|Maithili |mai|
+| 斯洛伐克文|Slovak |sk| | 昂加文|Angika |ang|
+| 斯洛文尼亚文|Slovenian |sl| | 孟加拉文|Bhojpuri |bho|
+| 阿尔巴尼亚文|Albanian |sq| | 摩揭陀文 |Magahi |mah|
+| 瑞典文|Swedish |sv| | 那格浦尔文|Nagpur |sck|
+| 西瓦希里文|Swahili |sw| | 尼瓦尔文|Newari |new|
+| 塔加洛文|Tagalog |tl| | 保加利亚文 |Goan Konkani|gom|
+| 土耳其文|Turkish |tr| | 沙特阿拉伯文|Saudi Arabia|sa|
+| 乌兹别克文|Uzbek |uz| | 阿瓦尔文|Avar |ava|
+| 越南文|Vietnamese |vi| | 阿瓦尔文|Avar |ava|
+| 蒙古文|Mongolian |mn| | 阿迪赫文|Adyghe |ady|
--- a/doc/doc_en/multi_languages_en.md
+++ b/doc/doc_en/multi_languages_en.md
@ -13,7 +13,7 @@ Among them, the English model supports the detection and recognition of uppercas
 letters and common punctuation, and the recognition of space characters is optimized:

 <div align="center">
-    <img src="../imgs_results/multi_lang/en_1.jpg" width="400" height="600">
+    <img src="../imgs_results/multi_lang/img_12.jpg" width="900" height="300">
 </div>

 The multilingual models cover Latin, Arabic, Traditional Chinese, Korean, Japanese, etc.:
@ -21,6 +21,8 @@ The multilingual models cover Latin, Arabic, Traditional Chinese, Korean, Japane
 <div align="center">
    <img src="../imgs_results/multi_lang/japan_2.jpg" width="600" height="300">
    <img src="../imgs_results/multi_lang/french_0.jpg" width="300" height="300">
+    <img src="../imgs_results/multi_lang/korean_0.jpg" width="500" height="300">
+    <img src="../imgs_results/multi_lang/arabic_0.jpg" width="300" height="300">
 </div>

 This document will briefly introduce how to use the multilingual model.
@ -31,14 +33,9 @@ This document will briefly introduce how to use the multilingual model.

 - [2 Quick Use](#Quick_Use)
    - [2.1 Command line operation](#Command_line_operation)
-     - [2.1.1 Prediction of the whole image](#bash_detection+recognition)
-     - [2.1.2 Recognition](#bash_Recognition)
-     - [2.1.3 Detection](#bash_detection)
    - [2.2 python script running](#python_Script_running)
-     - [2.2.1 Whole image prediction](#python_detection+recognition)
-     - [2.2.2 Recognition](#python_Recognition)
-     - [2.2.3 Detection](#python_detection)
 - [3 Custom Training](#Custom_Training)
+- [4 Inference and Deployment](#inference)
 - [4 Supported languages and abbreviations](#language_abbreviations)

 <a name="Install"></a>
@ -51,7 +48,7 @@ This document will briefly introduce how to use the multilingual model.
 pip install paddlepaddle

 # gpu
-pip instll paddlepaddle-gpu
+pip install paddlepaddle-gpu
 ```

 <a name="paddleocr_package_install"></a>
@ -89,7 +86,7 @@ The specific supported [language] (#language_abbreviations) can be viewed in the

 paddleocr --image_dir doc/imgs/japan_2.jpg --lang=japan
 ```
-![](https://raw.githubusercontent.com/PaddlePaddle/PaddleOCR/release/2.1/doc/imgs/japan_2.jpg)
+![](https://raw.githubusercontent.com/PaddlePaddle/PaddleOCR/release/2.0/doc/imgs/japan_2.jpg)

 The result is a list, each item contains a text box, text and recognition confidence
 ```text
@ -106,7 +103,7 @@ The result is a list, each item contains a text box, text and recognition confid
 paddleocr --image_dir doc/imgs_words/japan/1.jpg --det false --lang=japan
 ```

-![](https://raw.githubusercontent.com/PaddlePaddle/PaddleOCR/release/2.1/doc/imgs_words/japan/1.jpg)
+![](https://raw.githubusercontent.com/PaddlePaddle/PaddleOCR/release/2.0/doc/imgs_words/japan/1.jpg)

 The result is a tuple, which returns the recognition result and recognition confidence

@ -143,6 +140,9 @@ from paddleocr import PaddleOCR, draw_ocr
 ocr = PaddleOCR(lang="korean") # The model file will be downloaded automatically when executed for the first time
 img_path ='doc/imgs/korean_1.jpg'
 result = ocr.ocr(img_path)
+# Recognition and detection can be performed separately through parameter control
+# result = ocr.ocr(img_path, det=False)  Only perform recognition
+# result = ocr.ocr(img_path, rec=False)  Only perform detection
 # Print detection frame and recognition result
 for line in result:
    print(line)
@ -162,54 +162,6 @@ Visualization of results:
 ![](https://raw.githubusercontent.com/PaddlePaddle/PaddleOCR/release/2.1/doc/imgs_results/korean.jpg)


-* Recognition
-
-```
-from paddleocr import PaddleOCR
-ocr = PaddleOCR(lang="german")
-img_path ='PaddleOCR/doc/imgs_words/german/1.jpg'
-result = ocr.ocr(img_path, det=False, cls=True)
-for line in result:
-    print(line)
-```
-
-![](https://raw.githubusercontent.com/PaddlePaddle/PaddleOCR/release/2.1/doc/imgs_words/german/1.jpg)
-
-The result is a tuple, which only contains the recognition result and recognition confidence
-
-```
-('leider auch jetzt', 0.97538936)
-```
-
-* Detection
-
-```python
-from paddleocr import PaddleOCR, draw_ocr
-ocr = PaddleOCR() # need to run only once to download and load model into memory
-img_path ='PaddleOCR/doc/imgs_en/img_12.jpg'
-result = ocr.ocr(img_path, rec=False)
-for line in result:
-    print(line)
-
-# show result
-from PIL import Image
-
-image = Image.open(img_path).convert('RGB')
-im_show = draw_ocr(image, result, txts=None, scores=None, font_path='/path/to/PaddleOCR/doc/fonts/simfang.ttf')
-im_show = Image.fromarray(im_show)
-im_show.save('result.jpg')
-```
-The result is a list, each item contains only text boxes
-```bash
-[[26.0, 457.0], [137.0, 457.0], [137.0, 477.0], [26.0, 477.0]]
-[[25.0, 425.0], [372.0, 425.0], [372.0, 448.0], [25.0, 448.0]]
-[[128.0, 397.0], [273.0, 397.0], [273.0, 414.0], [128.0, 414.0]]
-......
-```
-
-Visualization of results:
-![](https://raw.githubusercontent.com/PaddlePaddle/PaddleOCR/release/2.1/doc/imgs_results/whl/12_det.jpg)
-
 ppocr also supports direction classification. For more usage methods, please refer to: [whl package instructions](https://github.com/PaddlePaddle/PaddleOCR/blob/release/2.0/doc/doc_ch/whl.md).

 <a name="Custom_training"></a>
@ -221,84 +173,61 @@ Modify the training data path, dictionary and other parameters.
 For specific data preparation and training process, please refer to: [Text Detection](../doc_en/detection_en.md), [Text Recognition](../doc_en/recognition_en.md), more functions such as predictive deployment,
 For functions such as data annotation, you can read the complete [Document Tutorial](../../README.md).

-<a name="language_abbreviation"></a>
-## 4 Support languages and abbreviations

-| Language  | Abbreviation |
-| ---  | --- |
-|chinese and english|ch|
-|english|en|
-|french|fr|
-|german|german|
-|japan|japan|
-|korean|korean|
-|chinese traditional |chinese_cht|
-| Italian |it|
-|Spanish |es|
-| Portuguese|pt|
-|Russia|ru|
-|Arabic|ar|
-|Hindi|hi|
-|Uyghur|ug|
-|Persian|fa|
-|Urdu|ur|
-| Serbian(latin) |rs_latin|
-|Occitan |oc|
-|Marathi|mr|
-|Nepali|ne|
-|Serbian(cyrillic)|rs_cyrillic|
-|Bulgarian |bg|
-|Ukranian|uk|
-|Belarusian|be|
-|Telugu |te|
-|Tamil |ta|
-|Afrikaans |af|
-|Azerbaijani    |az|
-|Bosnian|bs|
-|Czech|cs|
-|Welsh |cy|
-|Danish|da|
-|Estonian |et|
-|Irish |ga|
-|Croatian |hr|
-|Hungarian |hu|
-|Indonesian|id|
-|Icelandic|is|
-|Kurdish|ku|
-|Lithuanian |lt|
- |Latvian |lv|
-|Maori|mi|
-|Malay|ms|
-|Maltese |mt|
-|Dutch |nl|
-|Norwegian |no|
-|Polish |pl|
-|Romanian |ro|
-|Slovak |sk|
-|Slovenian |sl|
-|Albanian |sq|
-|Swedish |sv|
-|Swahili |sw|
-|Tagalog |tl|
-|Turkish |tr|
-|Uzbek |uz|
-|Vietnamese |vi|
-|Mongolian |mn|
-|Abaza |abq|
-|Adyghe |ady|
-|Kabardian |kbd|
-|Avar |ava|
-|Dargwa |dar|
-|Ingush |inh|
-|Lak |lbe|
-|Lezghian |lez|
-|Tabassaran |tab|
-|Bihari |bh|
-|Maithili |mai|
-|Angika |ang|
-|Bhojpuri |bho|
-|Magahi |mah|
-|Nagpur |sck|
-|Newari |new|
-|Goan Konkani|gom|
-|Saudi Arabia|sa|
+<a name="inference"></a>
+## 4 Inference and Deployment
+
+In addition to installing the whl package for quick forecasting,
+ppocr also provides a variety of forecasting deployment methods.
+If necessary, you can read related documents:
+
+- [Python Inference](./inference_en.md)
+- [C++ Inference](../../deploy/cpp_infer/readme_en.md)
+- [Serving](../../deploy/hubserving/readme_en.md)
+- [Mobile](https://github.com/PaddlePaddle/PaddleOCR/blob/develop/deploy/lite/readme_en.md)
+- [Benchmark](./benchmark_en.md)
+
+
+<a name="language_abbreviations"></a>
+## 5 Support languages and abbreviations
+
+| Language  | Abbreviation | | Language  | Abbreviation |
+| ---  | --- | --- | ---  | --- |
+|chinese and english|ch| |Arabic|ar|
+|english|en| |Hindi|hi|
+|french|fr| |Uyghur|ug|
+|german|german| |Persian|fa|
+|japan|japan| |Urdu|ur|
+|korean|korean| | Serbian(latin) |rs_latin|
+|chinese traditional |ch_tra| |Occitan |oc|
+| Italian |it| |Marathi|mr|
+|Spanish |es| |Nepali|ne|
+| Portuguese|pt| |Serbian(cyrillic)|rs_cyrillic|
+|Russia|ru||Bulgarian |bg|
+|Ukranian|uk| |Estonian |et|
+|Belarusian|be| |Irish |ga|
+|Telugu |te| |Croatian |hr|
+|Saudi Arabia|sa| |Hungarian |hu|
+|Tamil |ta| |Indonesian|id|
+|Afrikaans |af| |Icelandic|is|
+|Azerbaijani  |az||Kurdish|ku|
+|Bosnian|bs| |Lithuanian |lt|
+|Czech|cs| |Latvian |lv|
+|Welsh |cy| |Maori|mi|
+|Danish|da| |Malay|ms|
+|Maltese |mt| |Adyghe |ady|
+|Dutch |nl| |Kabardian |kbd|
+|Norwegian |no| |Avar |ava|
+|Polish |pl| |Dargwa |dar|
+|Romanian |ro| |Ingush |inh|
+|Slovak |sk| |Lak |lbe|
+|Slovenian |sl| |Lezghian |lez|
+|Albanian |sq| |Tabassaran |tab|
+|Swedish |sv| |Bihari |bh|
+|Swahili |sw| |Maithili |mai|
+|Tagalog |tl| |Angika |ang|
+|Turkish |tr| |Bhojpuri |bho|
+|Uzbek |uz| |Magahi |mah|
+|Vietnamese |vi| |Nagpur |sck|
+|Mongolian |mn| |Newari |new|
+|Abaza |abq| |Goan Konkani|gom|
--- a/doc/imgs_results/multi_lang/arabic_0.jpg
+++ b/doc/imgs_results/multi_lang/arabic_0.jpg
--- a/doc/imgs_results/multi_lang/img_12.jpg
+++ b/doc/imgs_results/multi_lang/img_12.jpg
--- a/doc/imgs_results/multi_lang/korean_0.jpg
+++ b/doc/imgs_results/multi_lang/korean_0.jpg
--- a/ppocr/modeling/backbones/rec_resnet_vd.py
+++ b/ppocr/modeling/backbones/rec_resnet_vd.py
@ -249,7 +249,7 @@ class ResNet(nn.Layer):
                            name=conv_name))
                    shortcut = True
                    self.block_list.append(bottleneck_block)
-                self.out_channels = num_filters[block]
+                self.out_channels = num_filters[block] * 4
        else:
            for block in range(len(depth)):
                shortcut = False
--- a/ppocr/postprocess/rec_postprocess.py
+++ b/ppocr/postprocess/rec_postprocess.py
@ -218,6 +218,7 @@ class SRNLabelDecode(BaseRecLabelDecode):
                 **kwargs):
        super(SRNLabelDecode, self).__init__(character_dict_path,
                                             character_type, use_space_char)
+        self.max_text_length = kwargs.get('max_text_length', 25)

    def __call__(self, preds, label=None, *args, **kwargs):
        pred = preds['predict']
@ -229,9 +230,9 @@ class SRNLabelDecode(BaseRecLabelDecode):
        preds_idx = np.argmax(pred, axis=1)
        preds_prob = np.max(pred, axis=1)

-        preds_idx = np.reshape(preds_idx, [-1, 25])
+        preds_idx = np.reshape(preds_idx, [-1, self.max_text_length])

-        preds_prob = np.reshape(preds_prob, [-1, 25])
+        preds_prob = np.reshape(preds_prob, [-1, self.max_text_length])

        text = self.decode(preds_idx, preds_prob)

--- a/ppocr/utils/save_load.py
+++ b/ppocr/utils/save_load.py
@ -121,7 +121,7 @@ def init_model(config, model, logger, optimizer=None, lr_scheduler=None):
    return best_model_dict


-def save_model(net,
+def save_model(model,
               optimizer,
               model_path,
               logger,
@ -133,7 +133,7 @@ def save_model(net,
    """
    _mkdir_if_not_exist(model_path, logger)
    model_prefix = os.path.join(model_path, prefix)
-    paddle.save(net.state_dict(), model_prefix + '.pdparams')
+    paddle.save(model.state_dict(), model_prefix + '.pdparams')
    paddle.save(optimizer.state_dict(), model_prefix + '.pdopt')

    # save metric and config
--- a/tools/export_model.py
+++ b/tools/export_model.py
@ -53,17 +53,19 @@ def main():
    save_path = '{}/inference'.format(config['Global']['save_inference_dir'])

    if config['Architecture']['algorithm'] == "SRN":
+        max_text_length = config['Architecture']['Head']['max_text_length']
        other_shape = [
            paddle.static.InputSpec(
                shape=[None, 1, 64, 256], dtype='float32'), [
                    paddle.static.InputSpec(
                        shape=[None, 256, 1],
                        dtype="int64"), paddle.static.InputSpec(
-                            shape=[None, 25, 1],
-                            dtype="int64"), paddle.static.InputSpec(
-                                shape=[None, 8, 25, 25], dtype="int64"),
+                            shape=[None, max_text_length, 1], dtype="int64"),
                    paddle.static.InputSpec(
-                        shape=[None, 8, 25, 25], dtype="int64")
+                        shape=[None, 8, max_text_length, max_text_length],
+                        dtype="int64"), paddle.static.InputSpec(
+                            shape=[None, 8, max_text_length, max_text_length],
+                            dtype="int64")
                ]
        ]
        model = to_static(model, input_spec=other_shape)
--- a/tools/program.py
+++ b/tools/program.py
@ -18,6 +18,7 @@ from __future__ import print_function

 import os
 import sys
+import platform
 import yaml
 import time
 import shutil
@ -159,6 +160,8 @@ def train(config,
    eval_batch_step = config['Global']['eval_batch_step']

    global_step = 0
+    if 'global_step' in pre_best_model_dict:
+        global_step = pre_best_model_dict['global_step']
    start_eval_step = 0
    if type(eval_batch_step) == list and len(eval_batch_step) >= 2:
        start_eval_step = eval_batch_step[0]
@ -196,8 +199,12 @@ def train(config,
        train_reader_cost = 0.0
        batch_sum = 0
        batch_start = time.time()
-        for idx, batch in enumerate(train_dataloader()):
+        max_iter = len(train_dataloader) - 1 if platform.system(
+        ) == "Windows" else len(train_dataloader)
+        for idx, batch in enumerate(train_dataloader):
            train_reader_cost += time.time() - batch_start
+            if idx >= max_iter:
+                break
            lr = optimizer.get_lr()
            images = batch[0]
            if use_srn:
@ -285,7 +292,8 @@ def train(config,
                        is_best=True,
                        prefix='best_accuracy',
                        best_model_dict=best_model_dict,
-                        epoch=epoch)
+                        epoch=epoch,
+                        global_step=global_step)
                best_str = 'best metric, {}'.format(', '.join([
                    '{}: {}'.format(k, v) for k, v in best_model_dict.items()
                ]))
@ -307,7 +315,8 @@ def train(config,
                is_best=False,
                prefix='latest',
                best_model_dict=best_model_dict,
-                epoch=epoch)
+                epoch=epoch,
+                global_step=global_step)
        if dist.get_rank() == 0 and epoch > 0 and epoch % save_epoch_step == 0:
            save_model(
                model,
@ -317,7 +326,8 @@ def train(config,
                is_best=False,
                prefix='iter_epoch_{}'.format(epoch),
                best_model_dict=best_model_dict,
-                epoch=epoch)
+                epoch=epoch,
+                global_step=global_step)
    best_str = 'best metric, {}'.format(', '.join(
        ['{}: {}'.format(k, v) for k, v in best_model_dict.items()]))
    logger.info(best_str)
@ -333,8 +343,10 @@ def eval(model, valid_dataloader, post_process_class, eval_class,
        total_frame = 0.0
        total_time = 0.0
        pbar = tqdm(total=len(valid_dataloader), desc='eval model:')
+        max_iter = len(valid_dataloader) - 1 if platform.system(
+        ) == "Windows" else len(valid_dataloader)
        for idx, batch in enumerate(valid_dataloader):
-            if idx >= len(valid_dataloader):
+            if idx >= max_iter:
                break
            images = batch[0]
            start = time.time()