Merge branch 'dygraph' of https://github.com/PaddlePaddle/PaddleOCR into doc

README_ch.md

@@ -8,9 +8,10 @@ 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。
-- 2021.1.26,28,29 PaddleOCR官方研发团队带来技术深入解读三日直播课，1月26日、28日、29日晚上19:30，[直播地址](https://live.bilibili.com/21689802)
-- 2021.1.21 更新多语言识别模型，目前支持语种超过27种，[多语言模型下载](./doc/doc_ch/models_list.md)，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)
+- 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)
 - 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。
 - 2020.11.25 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
@@ -74,11 +75,13 @@ PaddleOCR同时支持动态图与静态图两种编程范式
 ## 文档教程
 - [快速安装](./doc/doc_ch/installation.md)
 - [中文OCR模型快速使用](./doc/doc_ch/quickstart.md)
+- [多语言OCR模型快速使用](./doc/doc_ch/multi_languages.md)
 - [代码组织结构](./doc/doc_ch/tree.md)
 - 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)
+    - [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)
     - [文本识别](./doc/doc_ch/recognition.md)

configs/det/det_r50_vd_sast_icdar15.yml

@@ -14,12 +14,13 @@ Global:
   load_static_weights: True
   cal_metric_during_train: False
   pretrained_model: ./pretrain_models/ResNet50_vd_ssld_pretrained/
-  checkpoints: 
+  checkpoints:
   save_inference_dir:
   use_visualdl: False
-  infer_img: 
+  infer_img:
   save_res_path: ./output/sast_r50_vd_ic15/predicts_sast.txt
 
+
 Architecture:
   model_type: det
   algorithm: SAST

configs/e2e/e2e_r50_vd_pg.yml

+Global:
+  use_gpu: True
+  epoch_num: 600
+  log_smooth_window: 20
+  print_batch_step: 10
+  save_model_dir: ./output/pgnet_r50_vd_totaltext/
+  save_epoch_step: 10
+  # evaluation is run every 0 iterationss after the 1000th iteration
+  eval_batch_step: [ 0, 1000 ]
+  # 1. If pretrained_model is saved in static mode, such as classification pretrained model
+  #    from static branch, load_static_weights must be set as True.
+  # 2. If you want to finetune the pretrained models we provide in the docs,
+  #    you should set load_static_weights as False.
+  load_static_weights: False
+  cal_metric_during_train: False
+  pretrained_model:
+  checkpoints:
+  save_inference_dir:
+  use_visualdl: False
+  infer_img:
+  valid_set: totaltext # two mode: totaltext valid curved words, partvgg valid non-curved words
+  save_res_path: ./output/pgnet_r50_vd_totaltext/predicts_pgnet.txt
+  character_dict_path: ppocr/utils/ic15_dict.txt
+  character_type: EN
+  max_text_length: 50 # the max length in seq
+  max_text_nums: 30 # the max seq nums in a pic
+  tcl_len: 64
+
+Architecture:
+  model_type: e2e
+  algorithm: PGNet
+  Transform:
+  Backbone:
+    name: ResNet
+    layers: 50
+  Neck:
+    name: PGFPN
+  Head:
+    name: PGHead
+
+Loss:
+  name: PGLoss
+  tcl_bs: 64
+  max_text_length: 50 # the same as Global: max_text_length
+  max_text_nums: 30 # the same as Global：max_text_nums
+  pad_num: 36 # the length of dict for pad
+
+Optimizer:
+  name: Adam
+  beta1: 0.9
+  beta2: 0.999
+  lr:
+    learning_rate: 0.001
+  regularizer:
+    name: 'L2'
+    factor: 0
+
+
+PostProcess:
+  name: PGPostProcess
+  score_thresh: 0.5
+Metric:
+  name: E2EMetric
+  character_dict_path: ppocr/utils/ic15_dict.txt
+  main_indicator: f_score_e2e
+
+Train:
+  dataset:
+    name: PGDataSet
+    label_file_list: [.././train_data/total_text/train/]
+    ratio_list: [1.0]
+    data_format: icdar #two data format: icdar/textnet
+    transforms:
+      - DecodeImage: # load image
+          img_mode: BGR
+          channel_first: False
+      - PGProcessTrain:
+          batch_size: 14  # same as loader: batch_size_per_card
+          min_crop_size: 24
+          min_text_size: 4
+          max_text_size: 512
+      - KeepKeys:
+          keep_keys: [ 'images', 'tcl_maps', 'tcl_label_maps', 'border_maps','direction_maps', 'training_masks', 'label_list', 'pos_list', 'pos_mask' ] # dataloader will return list in this order
+  loader:
+    shuffle: True
+    drop_last: True
+    batch_size_per_card: 14
+    num_workers: 16
+
+Eval:
+  dataset:
+    name: PGDataSet
+    data_dir: ./train_data/
+    label_file_list: [./train_data/total_text/test/]
+    transforms:
+      - DecodeImage: # load image
+          img_mode: RGB
+          channel_first: False
+      - E2ELabelEncode:
+      - E2EResizeForTest:
+          max_side_len: 768
+      - NormalizeImage:
+          scale: 1./255.
+          mean: [ 0.485, 0.456, 0.406 ]
+          std: [ 0.229, 0.224, 0.225 ]
+          order: 'hwc'
+      - ToCHWImage:
+      - KeepKeys:
+          keep_keys: [ 'image', 'shape', 'polys', 'strs', 'tags' ]
+  loader:
+    shuffle: False
+    drop_last: False
+    batch_size_per_card: 1 # must be 1
+    num_workers: 2
\ No newline at end of file

doc/doc_ch/inference.md

@@ -12,7 +12,8 @@ inference 模型（`paddle.jit.save`保存的模型）
 - [一、训练模型转inference模型](#训练模型转inference模型)
     - [检测模型转inference模型](#检测模型转inference模型)
     - [识别模型转inference模型](#识别模型转inference模型)  
-    - [方向分类模型转inference模型](#方向分类模型转inference模型)  
+    - [方向分类模型转inference模型](#方向分类模型转inference模型)
+    - [端到端模型转inference模型](#端到端模型转inference模型)
 
 - [二、文本检测模型推理](#文本检测模型推理)
     - [1. 超轻量中文检测模型推理](#超轻量中文检测模型推理)
@@ -27,10 +28,13 @@ inference 模型（`paddle.jit.save`保存的模型）
     - [4. 自定义文本识别字典的推理](#自定义文本识别字典的推理)
     - [5. 多语言模型的推理](#多语言模型的推理)
 
-- [四、方向分类模型推理](#方向识别模型推理)
+- [四、端到端模型推理](#端到端模型推理)
+    - [1. PGNet端到端模型推理](#PGNet端到端模型推理)
+
+- [五、方向分类模型推理](#方向识别模型推理)
     - [1. 方向分类模型推理](#方向分类模型推理)
 
-- [五、文本检测、方向分类和文字识别串联推理](#文本检测、方向分类和文字识别串联推理)
+- [六、文本检测、方向分类和文字识别串联推理](#文本检测、方向分类和文字识别串联推理)
     - [1. 超轻量中文OCR模型推理](#超轻量中文OCR模型推理)
     - [2. 其他模型推理](#其他模型推理)
 
@@ -118,6 +122,32 @@ python3 tools/export_model.py -c configs/cls/cls_mv3.yml -o Global.pretrained_mo
     ├── inference.pdiparams.info    # 分类inference模型的参数信息，可忽略
     └── inference.pdmodel           # 分类inference模型的program文件
 ```
+<a name="端到端模型转inference模型"></a>
+### 端到端模型转inference模型
+
+下载端到端模型：
+```
+wget -P ./ch_lite/ https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_cls_train.tar && tar xf ./ch_lite/ch_ppocr_mobile_v2.0_cls_train.tar -C ./ch_lite/
+```
+
+端到端模型转inference模型与检测的方式相同，如下：
+```
+# -c 后面设置训练算法的yml配置文件
+# -o 配置可选参数
+# Global.pretrained_model 参数设置待转换的训练模型地址，不用添加文件后缀 .pdmodel，.pdopt或.pdparams。
+# Global.load_static_weights 参数需要设置为 False。
+# Global.save_inference_dir参数设置转换的模型将保存的地址。
+
+python3 tools/export_model.py -c configs/e2e/e2e_r50_vd_pg.yml -o Global.pretrained_model=./ch_lite/ch_ppocr_mobile_v2.0_cls_train/best_accuracy Global.load_static_weights=False Global.save_inference_dir=./inference/e2e/
+```
+
+转换成功后，在目录下有三个文件：
+```
+/inference/e2e/
+    ├── inference.pdiparams         # 分类inference模型的参数文件
+    ├── inference.pdiparams.info    # 分类inference模型的参数信息，可忽略
+    └── inference.pdmodel           # 分类inference模型的program文件
+```
 
 <a name="文本检测模型推理"></a>
 ## 二、文本检测模型推理
@@ -332,8 +362,38 @@ python3 tools/infer/predict_rec.py --image_dir="./doc/imgs_words/korean/1.jpg" -
 Predicts of ./doc/imgs_words/korean/1.jpg:('바탕으로', 0.9948904)
 ```
 
+<a name="端到端模型推理"></a>
+## 四、端到端模型推理
+
+端到端模型推理，默认使用PGNet模型的配置参数。当不使用PGNet模型时，在推理时，需要通过传入相应的参数进行算法适配，细节参考下文。
+<a name="PGNet端到端模型推理"></a>
+### 1. PGNet端到端模型推理
+#### (1). 四边形文本检测模型（ICDAR2015）  
+首先将PGNet端到端训练过程中保存的模型，转换成inference model。以基于Resnet50_vd骨干网络，在ICDAR2015英文数据集训练的模型为例([模型下载地址](https://paddleocr.bj.bcebos.com/dygraph_v2.0/pgnet/en_server_pgnetA.tar))，可以使用如下命令进行转换：
+```
+python3 tools/export_model.py -c configs/e2e/e2e_r50_vd_pg.yml -o Global.pretrained_model=./en_server_pgnetA/iter_epoch_450 Global.load_static_weights=False Global.save_inference_dir=./inference/e2e
+```
+**PGNet端到端模型推理，需要设置参数`--e2e_algorithm="PGNet"`**，可以执行如下命令：
+```
+python3 tools/infer/predict_e2e.py --e2e_algorithm="PGNet" --image_dir="./doc/imgs_en/img_10.jpg" --e2e_model_dir="./inference/e2e/"  --e2e_pgnet_polygon=False
+```
+可视化文本检测结果默认保存到`./inference_results`文件夹里面，结果文件的名称前缀为'e2e_res'。结果示例如下：
+
+![](../imgs_results/e2e_res_img_10_pgnet.jpg)
+
+#### (2). 弯曲文本检测模型（Total-Text）  
+和四边形文本检测模型共用一个推理模型
+**PGNet端到端模型推理，需要设置参数`--e2e_algorithm="PGNet"`，同时，还需要增加参数`--e2e_pgnet_polygon=True`，**可以执行如下命令：
+```
+python3.7 tools/infer/predict_e2e.py --e2e_algorithm="PGNet" --image_dir="./doc/imgs_en/img623.jpg" --e2e_model_dir="./inference/e2e/" --e2e_pgnet_polygon=True
+```
+可视化文本端到端结果默认保存到`./inference_results`文件夹里面，结果文件的名称前缀为'e2e_res'。结果示例如下：
+
+![](../imgs_results/e2e_res_img623_pgnet.jpg)
+
+
 <a name="方向分类模型推理"></a>
-## 四、方向分类模型推理
+## 五、方向分类模型推理
 
 下面将介绍方向分类模型推理。
 
@@ -358,7 +418,7 @@ Predicts of ./doc/imgs_words/ch/word_4.jpg:['0', 0.9999982]
 ```
 
 <a name="文本检测、方向分类和文字识别串联推理"></a>
-## 五、文本检测、方向分类和文字识别串联推理
+## 六、文本检测、方向分类和文字识别串联推理
 <a name="超轻量中文OCR模型推理"></a>
 ### 1. 超轻量中文OCR模型推理
 

doc/doc_ch/pgnet.md

+# 端对端OCR算法-PGNet
+- [一、简介](#简介)
+- [二、环境配置](#环境配置)
+- [三、快速使用](#快速使用)
+- [四、模型训练、评估、推理](#快速训练)
+
+<a name="简介"></a>
+## 一、简介
+OCR算法可以分为两阶段算法和端对端的算法。二阶段OCR算法一般分为两个部分，文本检测和文本识别算法，文件检测算法从图像中得到文本行的检测框，然后识别算法去识别文本框中的内容。而端对端OCR算法可以在一个算法中完成文字检测和文字识别，其基本思想是设计一个同时具有检测单元和识别模块的模型，共享其中两者的CNN特征，并联合训练。由于一个算法即可完成文字识别，端对端模型更小，速度更快。
+
+### PGNet算法介绍
+近些年来，端对端OCR算法得到了良好的发展，包括MaskTextSpotter系列、TextSnake、TextDragon、PGNet系列等算法。在这些算法中，PGNet算法具备其他算法不具备的优势，包括：
+- 设计PGNet loss指导训练，不需要字符级别的标注
+- 不需要NMS和ROI相关操作，加速预测
+- 提出预测文本行内的阅读顺序模块；
+- 提出基于图的修正模块（GRM）来进一步提高模型识别性能
+- 精度更高，预测速度更快
+
+PGNet算法细节详见[论文](https://www.aaai.org/AAAI21Papers/AAAI-2885.WangP.pdf)， 算法原理图如下所示：
+![](../pgnet_framework.png)
+输入图像经过特征提取送入四个分支，分别是：文本边缘偏移量预测TBO模块，文本中心线预测TCL模块，文本方向偏移量预测TDO模块，以及文本字符分类图预测TCC模块。
+其中TBO以及TCL的输出经过后处理后可以得到文本的检测结果，TCL、TDO、TCC负责文本识别。
+其检测识别效果图如下：
+![](../imgs_results/e2e_res_img293_pgnet.png)
+![](../imgs_results/e2e_res_img295_pgnet.png)
+
+<a name="环境配置"></a>
+## 二、环境配置
+请先参考[快速安装](./installation.md)配置PaddleOCR运行环境。
+
+<a name="快速使用"></a>
+## 三、快速使用
+### inference模型下载
+本节以训练好的端到端模型为例，快速使用模型预测，首先下载训练好的端到端inference模型[下载地址](https://paddleocr.bj.bcebos.com/dygraph_v2.0/pgnet/e2e_server_pgnetA_infer.tar)
+```
+mkdir inference && cd inference
+# 下载英文端到端模型并解压
+wget https://paddleocr.bj.bcebos.com/dygraph_v2.0/pgnet/e2e_server_pgnetA_infer.tar && tar xf e2e_server_pgnetA_infer.tar
+```
+* windows 环境下如果没有安装wget,下载模型时可将链接复制到浏览器中下载，并解压放置在相应目录下
+
+解压完毕后应有如下文件结构：
+```
+├── e2e_server_pgnetA_infer
+│   ├── inference.pdiparams
+│   ├── inference.pdiparams.info
+│   └── inference.pdmodel
+```
+### 单张图像或者图像集合预测
+```bash
+# 预测image_dir指定的单张图像
+python3 tools/infer/predict_e2e.py --e2e_algorithm="PGNet" --image_dir="./doc/imgs_en/img623.jpg" --e2e_model_dir="./inference/e2e/" --e2e_pgnet_polygon=True
+
+# 预测image_dir指定的图像集合
+python3 tools/infer/predict_e2e.py --e2e_algorithm="PGNet" --image_dir="./doc/imgs_en/" --e2e_model_dir="./inference/e2e/" --e2e_pgnet_polygon=True
+
+# 如果想使用CPU进行预测，需设置use_gpu参数为False
+python3 tools/infer/predict_e2e.py --e2e_algorithm="PGNet" --image_dir="./doc/imgs_en/img623.jpg" --e2e_model_dir="./inference/e2e/" --e2e_pgnet_polygon=True --use_gpu=False
+```
+### 可视化结果
+可视化文本检测结果默认保存到./inference_results文件夹里面，结果文件的名称前缀为'e2e_res'。结果示例如下：
+![](../imgs_results/e2e_res_img623_pgnet.jpg)
+
+<a name="快速训练"></a>
+## 四、模型训练、评估、推理
+本节以totaltext数据集为例，介绍PaddleOCR中端到端模型的训练、评估与测试。
+
+###  准备数据
+下载解压[totaltext](https://github.com/cs-chan/Total-Text-Dataset/blob/master/Dataset/README.md)数据集到PaddleOCR/train_data/目录，数据集组织结构：
+```
+/PaddleOCR/train_data/total_text/train/
+  |- rgb/            # total_text数据集的训练数据
+      |- gt_0.png
+      | ...  
+  |- total_text.txt  # total_text数据集的训练标注
+```
+
+total_text.txt标注文件格式如下，文件名和标注信息中间用"\t"分隔：
+```
+" 图像文件名                    json.dumps编码的图像标注信息"
+rgb/gt_0.png    [{"transcription": "EST", "points": [[1004.0,689.0],[1019.0,698.0],[1034.0,708.0],[1049.0,718.0],[1064.0,728.0],[1079.0,738.0],[1095.0,748.0],[1094.0,774.0],[1079.0,765.0],[1065.0,756.0],[1050.0,747.0],[1036.0,738.0],[1021.0,729.0],[1007.0,721.0]]}, {...}]
+```
+json.dumps编码前的图像标注信息是包含多个字典的list，字典中的 `points` 表示文本框的四个点的坐标(x, y)，从左上角的点开始顺时针排列。
+`transcription` 表示当前文本框的文字，**当其内容为“###”时，表示该文本框无效，在训练时会跳过。**
+如果您想在其他数据集上训练，可以按照上述形式构建标注文件。
+
+### 启动训练
+
+PGNet训练分为两个步骤：step1: 在合成数据上训练，得到预训练模型，此时模型精度依然较低；step2: 加载预训练模型，在totaltext数据集上训练；为快速训练，我们直接提供了step1的预训练模型。
+```shell
+cd PaddleOCR/
+下载step1 预训练模型
+wget -P ./pretrain_models/ https://paddleocr.bj.bcebos.com/dygraph_v2.0/pgnet/train_step1.tar
+可以得到以下的文件格式
+./pretrain_models/train_step1/
+  └─ best_accuracy.pdopt
+  └─ best_accuracy.states
+  └─ best_accuracy.pdparams
+```
+*如果您安装的是cpu版本，请将配置文件中的 `use_gpu` 字段修改为false*
+
+```shell
+# 单机单卡训练 e2e 模型
+python3 tools/train.py -c configs/e2e/e2e_r50_vd_pg.yml -o Global.pretrained_model=./pretrain_models/train_step1/best_accuracy Global.load_static_weights=False
+# 单机多卡训练，通过 --gpus 参数设置使用的GPU ID
+python3 -m paddle.distributed.launch --gpus '0,1,2,3' tools/train.py -c configs/e2e/e2e_r50_vd_pg.yml -o Global.pretrained_model=./pretrain_models/train_step1/best_accuracy  Global.load_static_weights=False
+```
+
+上述指令中，通过-c 选择训练使用configs/e2e/e2e_r50_vd_pg.yml配置文件。
+有关配置文件的详细解释，请参考[链接](./config.md)。
+
+您也可以通过-o参数在不需要修改yml文件的情况下，改变训练的参数，比如，调整训练的学习率为0.0001
+```shell
+python3 tools/train.py -c configs/e2e/e2e_r50_vd_pg.yml -o Optimizer.base_lr=0.0001
+```
+
+#### 断点训练
+如果训练程序中断，如果希望加载训练中断的模型从而恢复训练，可以通过指定Global.checkpoints指定要加载的模型路径：
+```shell
+python3 tools/train.py -c configs/e2e/e2e_r50_vd_pg.yml -o Global.checkpoints=./your/trained/model
+```
+
+**注意**：`Global.checkpoints`的优先级高于`Global.pretrain_weights`的优先级，即同时指定两个参数时，优先加载`Global.checkpoints`指定的模型，如果`Global.checkpoints`指定的模型路径有误，会加载`Global.pretrain_weights`指定的模型。
+
+PaddleOCR计算三个OCR端到端相关的指标，分别是：Precision、Recall、Hmean。
+
+运行如下代码，根据配置文件`e2e_r50_vd_pg.yml`中`save_res_path`指定的测试集检测结果文件，计算评估指标。
+
+评估时设置后处理参数`max_side_len=768`，使用不同数据集、不同模型训练，可调整参数进行优化
+训练中模型参数默认保存在`Global.save_model_dir`目录下。在评估指标时，需要设置`Global.checkpoints`指向保存的参数文件。
+```shell
+python3 tools/eval.py -c configs/e2e/e2e_r50_vd_pg.yml  -o Global.checkpoints="{path/to/weights}/best_accuracy"
+```
+
+### 模型预测
+测试单张图像的端到端识别效果
+```shell
+python3 tools/infer_e2e.py -c configs/e2e/e2e_r50_vd_pg.yml -o Global.infer_img="./doc/imgs_en/img_10.jpg" Global.pretrained_model="./output/det_db/best_accuracy" Global.load_static_weights=false
+```
+
+测试文件夹下所有图像的端到端识别效果
+```shell
+python3 tools/infer_e2e.py -c configs/e2e/e2e_r50_vd_pg.yml -o Global.infer_img="./doc/imgs_en/" Global.pretrained_model="./output/det_db/best_accuracy" Global.load_static_weights=false
+```
+
+### 预测推理
+#### (1).四边形文本检测模型（ICDAR2015）  
+首先将PGNet端到端训练过程中保存的模型，转换成inference model。以基于Resnet50_vd骨干网络，以英文数据集训练的模型为例[模型下载地址](https://paddleocr.bj.bcebos.com/dygraph_v2.0/pgnet/en_server_pgnetA.tar) ，可以使用如下命令进行转换：
+```
+wget https://paddleocr.bj.bcebos.com/dygraph_v2.0/pgnet/en_server_pgnetA.tar && tar xf en_server_pgnetA.tar
+python3 tools/export_model.py -c configs/e2e/e2e_r50_vd_pg.yml -o Global.pretrained_model=./en_server_pgnetA/iter_epoch_450 Global.load_static_weights=False Global.save_inference_dir=./inference/e2e
+```
+**PGNet端到端模型推理，需要设置参数`--e2e_algorithm="PGNet"`**，可以执行如下命令：
+```
+python3 tools/infer/predict_e2e.py --e2e_algorithm="PGNet" --image_dir="./doc/imgs_en/img_10.jpg" --e2e_model_dir="./inference/e2e/"  --e2e_pgnet_polygon=False
+```
+可视化文本检测结果默认保存到`./inference_results`文件夹里面，结果文件的名称前缀为'e2e_res'。结果示例如下：
+
+![](../imgs_results/e2e_res_img_10_pgnet.jpg)
+
+#### (2).弯曲文本检测模型（Total-Text）
+对于弯曲文本样例
+
+**PGNet端到端模型推理，需要设置参数`--e2e_algorithm="PGNet"`，同时，还需要增加参数`--e2e_pgnet_polygon=True`，**可以执行如下命令：
+```
+python3 tools/infer/predict_e2e.py --e2e_algorithm="PGNet" --image_dir="./doc/imgs_en/img623.jpg" --e2e_model_dir="./inference/e2e/" --e2e_pgnet_polygon=True
+```
+可视化文本端到端结果默认保存到`./inference_results`文件夹里面，结果文件的名称前缀为'e2e_res'。结果示例如下：
+
+![](../imgs_results/e2e_res_img623_pgnet.jpg)

ppocr/data/__init__.py

@@ -34,6 +34,7 @@ import paddle.distributed as dist
 from ppocr.data.imaug import transform, create_operators
 from ppocr.data.simple_dataset import SimpleDataSet
 from ppocr.data.lmdb_dataset import LMDBDataSet
+from ppocr.data.pgnet_dataset import PGDataSet
 
 __all__ = ['build_dataloader', 'transform', 'create_operators']
 
@@ -54,7 +55,7 @@ signal.signal(signal.SIGTERM, term_mp)
 def build_dataloader(config, mode, device, logger, seed=None):
     config = copy.deepcopy(config)
 
-    support_dict = ['SimpleDataSet', 'LMDBDataSet']
+    support_dict = ['SimpleDataSet', 'LMDBDataSet', 'PGDataSet']
     module_name = config[mode]['dataset']['name']
     assert module_name in support_dict, Exception(
         'DataSet only support {}'.format(support_dict))
@@ -72,14 +73,14 @@ def build_dataloader(config, mode, device, logger, seed=None):
     else:
         use_shared_memory = True
     if mode == "Train":
-        #Distribute data to multiple cards
+        # Distribute data to multiple cards
         batch_sampler = DistributedBatchSampler(
             dataset=dataset,
             batch_size=batch_size,
             shuffle=shuffle,
             drop_last=drop_last)
     else:
-        #Distribute data to single card
+        # Distribute data to single card
         batch_sampler = BatchSampler(
             dataset=dataset,
             batch_size=batch_size,

ppocr/data/imaug/__init__.py

@@ -28,6 +28,7 @@ from .label_ops import *
 
 from .east_process import *
 from .sast_process import *
+from .pg_process import *
 
 
 def transform(data, ops=None):

ppocr/data/imaug/label_ops.py

@@ -187,6 +187,34 @@ class CTCLabelEncode(BaseRecLabelEncode):
         return dict_character
 
 
+class E2ELabelEncode(BaseRecLabelEncode):
+    def __init__(self,
+                 max_text_length,
+                 character_dict_path=None,
+                 character_type='EN',
+                 use_space_char=False,
+                 **kwargs):
+        super(E2ELabelEncode,
+              self).__init__(max_text_length, character_dict_path,
+                             character_type, use_space_char)
+        self.pad_num = len(self.dict)  # the length to pad
+
+    def __call__(self, data):
+        text_label_index_list, temp_text = [], []
+        texts = data['strs']
+        for text in texts:
+            text = text.lower()
+            temp_text = []
+            for c_ in text:
+                if c_ in self.dict:
+                    temp_text.append(self.dict[c_])
+            temp_text = temp_text + [self.pad_num] * (self.max_text_len -
+                                                      len(temp_text))
+            text_label_index_list.append(temp_text)
+        data['strs'] = np.array(text_label_index_list)
+        return data
+
+
 class AttnLabelEncode(BaseRecLabelEncode):
     """ Convert between text-label and text-index """
 

ppocr/data/imaug/operators.py

@@ -197,7 +197,6 @@ class DetResizeForTest(object):
             sys.exit(0)
         ratio_h = resize_h / float(h)
         ratio_w = resize_w / float(w)
-        # return img, np.array([h, w])
         return img, [ratio_h, ratio_w]
 
     def resize_image_type2(self, img):
@@ -206,7 +205,6 @@ class DetResizeForTest(object):
         resize_w = w
         resize_h = h
 
-        # Fix the longer side
         if resize_h > resize_w:
             ratio = float(self.resize_long) / resize_h
         else:
@@ -223,3 +221,72 @@ class DetResizeForTest(object):
         ratio_w = resize_w / float(w)
 
         return img, [ratio_h, ratio_w]
+
+
+class E2EResizeForTest(object):
+    def __init__(self, **kwargs):
+        super(E2EResizeForTest, self).__init__()
+        self.max_side_len = kwargs['max_side_len']
+        self.valid_set = kwargs['valid_set']
+
+    def __call__(self, data):
+        img = data['image']
+        src_h, src_w, _ = img.shape
+        if self.valid_set == 'totaltext':
+            im_resized, [ratio_h, ratio_w] = self.resize_image_for_totaltext(
+                img, max_side_len=self.max_side_len)
+        else:
+            im_resized, (ratio_h, ratio_w) = self.resize_image(
+                img, max_side_len=self.max_side_len)
+        data['image'] = im_resized
+        data['shape'] = np.array([src_h, src_w, ratio_h, ratio_w])
+        return data
+
+    def resize_image_for_totaltext(self, im, max_side_len=512):
+
+        h, w, _ = im.shape
+        resize_w = w
+        resize_h = h
+        ratio = 1.25
+        if h * ratio > max_side_len:
+            ratio = float(max_side_len) / resize_h
+        resize_h = int(resize_h * ratio)
+        resize_w = int(resize_w * ratio)
+
+        max_stride = 128
+        resize_h = (resize_h + max_stride - 1) // max_stride * max_stride
+        resize_w = (resize_w + max_stride - 1) // max_stride * max_stride
+        im = cv2.resize(im, (int(resize_w), int(resize_h)))
+        ratio_h = resize_h / float(h)
+        ratio_w = resize_w / float(w)
+        return im, (ratio_h, ratio_w)
+
+    def resize_image(self, im, max_side_len=512):
+        """
+        resize image to a size multiple of max_stride which is required by the network
+        :param im: the resized image
+        :param max_side_len: limit of max image size to avoid out of memory in gpu
+        :return: the resized image and the resize ratio
+        """
+        h, w, _ = im.shape
+
+        resize_w = w
+        resize_h = h
+
+        # Fix the longer side
+        if resize_h > resize_w:
+            ratio = float(max_side_len) / resize_h
+        else:
+            ratio = float(max_side_len) / resize_w
+
+        resize_h = int(resize_h * ratio)
+        resize_w = int(resize_w * ratio)
+
+        max_stride = 128
+        resize_h = (resize_h + max_stride - 1) // max_stride * max_stride
+        resize_w = (resize_w + max_stride - 1) // max_stride * max_stride
+        im = cv2.resize(im, (int(resize_w), int(resize_h)))
+        ratio_h = resize_h / float(h)
+        ratio_w = resize_w / float(w)
+
+        return im, (ratio_h, ratio_w)

ppocr/data/pgnet_dataset.py

+# copyright (c) 2021 PaddlePaddle Authors. All Rights Reserve.
+#
+# 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.
+import numpy as np
+import os
+from paddle.io import Dataset
+from .imaug import transform, create_operators
+import random
+
+
+class PGDataSet(Dataset):
+    def __init__(self, config, mode, logger, seed=None):
+        super(PGDataSet, self).__init__()
+
+        self.logger = logger
+        self.seed = seed
+        self.mode = mode
+        global_config = config['Global']
+        dataset_config = config[mode]['dataset']
+        loader_config = config[mode]['loader']
+
+        label_file_list = dataset_config.pop('label_file_list')
+        data_source_num = len(label_file_list)
+        ratio_list = dataset_config.get("ratio_list", [1.0])
+        if isinstance(ratio_list, (float, int)):
+            ratio_list = [float(ratio_list)] * int(data_source_num)
+        self.data_format = dataset_config.get('data_format', 'icdar')
+        assert len(
+            ratio_list
+        ) == data_source_num, "The length of ratio_list should be the same as the file_list."
+        self.do_shuffle = loader_config['shuffle']
+
+        logger.info("Initialize indexs of datasets:%s" % label_file_list)
+        self.data_lines = self.get_image_info_list(label_file_list, ratio_list,
+                                                   self.data_format)
+        self.data_idx_order_list = list(range(len(self.data_lines)))
+        if mode.lower() == "train":
+            self.shuffle_data_random()
+
+        self.ops = create_operators(dataset_config['transforms'], global_config)
+
+    def shuffle_data_random(self):
+        if self.do_shuffle:
+            random.seed(self.seed)
+            random.shuffle(self.data_lines)
+        return
+
+    def extract_polys(self, poly_txt_path):
+        """
+        Read text_polys, txt_tags, txts from give txt file.
+        """
+        text_polys, txt_tags, txts = [], [], []
+        with open(poly_txt_path) as f:
+            for line in f.readlines():
+                poly_str, txt = line.strip().split('\t')
+                poly = list(map(float, poly_str.split(',')))
+                if self.mode.lower() == "eval":
+                    while len(poly) < 100:
+                        poly.append(-1)
+                text_polys.append(
+                    np.array(
+                        poly, dtype=np.float32).reshape(-1, 2))
+                txts.append(txt)
+                txt_tags.append(txt == '###')
+
+        return np.array(list(map(np.array, text_polys))), \
+               np.array(txt_tags, dtype=np.bool), txts
+
+    def extract_info_textnet(self, im_fn, img_dir=''):
+        """
+        Extract information from line in textnet format.
+        """
+        info_list = im_fn.split('\t')
+        img_path = ''
+        for ext in [
+                'jpg', 'bmp', 'png', 'jpeg', 'rgb', 'tif', 'tiff', 'gif', 'JPG'
+        ]:
+            if os.path.exists(os.path.join(img_dir, info_list[0] + "." + ext)):
+                img_path = os.path.join(img_dir, info_list[0] + "." + ext)
+                break
+
+        if img_path == '':
+            print('Image {0} NOT found in {1}, and it will be ignored.'.format(
+                info_list[0], img_dir))
+
+        nBox = (len(info_list) - 1) // 9
+        wordBBs, txts, txt_tags = [], [], []
+        for n in range(0, nBox):
+            wordBB = list(map(float, info_list[n * 9 + 1:(n + 1) * 9]))
+            txt = info_list[(n + 1) * 9]
+            wordBBs.append([[wordBB[0], wordBB[1]], [wordBB[2], wordBB[3]],
+                            [wordBB[4], wordBB[5]], [wordBB[6], wordBB[7]]])
+            txts.append(txt)
+            if txt == '###':
+                txt_tags.append(True)
+            else:
+                txt_tags.append(False)
+        return img_path, np.array(wordBBs, dtype=np.float32), txt_tags, txts
+
+    def get_image_info_list(self, file_list, ratio_list, data_format='textnet'):
+        if isinstance(file_list, str):
+            file_list = [file_list]
+        data_lines = []
+        for idx, data_source in enumerate(file_list):
+            image_files = []
+            if data_format == 'icdar':
+                image_files = [(data_source, x) for x in
+                               os.listdir(os.path.join(data_source, 'rgb'))
+                               if x.split('.')[-1] in [
+                                   'jpg', 'bmp', 'png', 'jpeg', 'rgb', 'tif',
+                                   'tiff', 'gif', 'JPG'
+                               ]]
+            elif data_format == 'textnet':
+                with open(data_source) as f:
+                    image_files = [(data_source, x.strip())
+                                   for x in f.readlines()]
+            else:
+                print("Unrecognized data format...")
+                exit(-1)
+            random.seed(self.seed)
+            image_files = random.sample(
+                image_files, round(len(image_files) * ratio_list[idx]))
+            data_lines.extend(image_files)
+        return data_lines
+
+    def __getitem__(self, idx):
+        file_idx = self.data_idx_order_list[idx]
+        data_path, data_line = self.data_lines[file_idx]
+        try:
+            if self.data_format == 'icdar':
+                im_path = os.path.join(data_path, 'rgb', data_line)
+                if self.mode.lower() == "eval":
+                    poly_path = os.path.join(data_path, 'poly_gt',
+                                             data_line.split('.')[0] + '.txt')
+                else:
+                    poly_path = os.path.join(data_path, 'poly',
+                                             data_line.split('.')[0] + '.txt')
+                text_polys, text_tags, text_strs = self.extract_polys(poly_path)
+            else:
+                image_dir = os.path.join(os.path.dirname(data_path), 'image')
+                im_path, text_polys, text_tags, text_strs = self.extract_info_textnet(
+                    data_line, image_dir)
+
+            data = {
+                'img_path': im_path,
+                'polys': text_polys,
+                'tags': text_tags,
+                'strs': text_strs
+            }
+            with open(data['img_path'], 'rb') as f:
+                img = f.read()
+                data['image'] = img
+            outs = transform(data, self.ops)
+
+        except Exception as e:
+            self.logger.error(
+                "When parsing line {}, error happened with msg: {}".format(
+                    self.data_idx_order_list[idx], e))
+            outs = None
+        if outs is None:
+            return self.__getitem__(np.random.randint(self.__len__()))
+        return outs
+
+    def __len__(self):
+        return len(self.data_idx_order_list)

ppocr/losses/__init__.py

@@ -29,10 +29,11 @@ def build_loss(config):
     # cls loss
     from .cls_loss import ClsLoss
 
+    # e2e loss
+    from .e2e_pg_loss import PGLoss
     support_dict = [
         'DBLoss', 'EASTLoss', 'SASTLoss', 'CTCLoss', 'ClsLoss', 'AttentionLoss',
-        'SRNLoss'
-    ]
+        'SRNLoss', 'PGLoss']
 
     config = copy.deepcopy(config)
     module_name = config.pop('name')

ppocr/losses/e2e_pg_loss.py

+# copyright (c) 2021 PaddlePaddle Authors. All Rights Reserve.
+#
+# 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.
+
+from __future__ import absolute_import
+from __future__ import division
+from __future__ import print_function
+
+from paddle import nn
+import paddle
+
+from .det_basic_loss import DiceLoss
+from ppocr.utils.e2e_utils.extract_batchsize import pre_process
+
+
+class PGLoss(nn.Layer):
+    def __init__(self,
+                 tcl_bs,
+                 max_text_length,
+                 max_text_nums,
+                 pad_num,
+                 eps=1e-6,
+                 **kwargs):
+        super(PGLoss, self).__init__()
+        self.tcl_bs = tcl_bs
+        self.max_text_nums = max_text_nums
+        self.max_text_length = max_text_length
+        self.pad_num = pad_num
+        self.dice_loss = DiceLoss(eps=eps)
+
+    def border_loss(self, f_border, l_border, l_score, l_mask):
+        l_border_split, l_border_norm = paddle.tensor.split(
+            l_border, num_or_sections=[4, 1], axis=1)
+        f_border_split = f_border
+        b, c, h, w = l_border_norm.shape
+        l_border_norm_split = paddle.expand(
+            x=l_border_norm, shape=[b, 4 * c, h, w])
+        b, c, h, w = l_score.shape
+        l_border_score = paddle.expand(x=l_score, shape=[b, 4 * c, h, w])
+        b, c, h, w = l_mask.shape
+        l_border_mask = paddle.expand(x=l_mask, shape=[b, 4 * c, h, w])
+        border_diff = l_border_split - f_border_split
+        abs_border_diff = paddle.abs(border_diff)
+        border_sign = abs_border_diff < 1.0
+        border_sign = paddle.cast(border_sign, dtype='float32')
+        border_sign.stop_gradient = True
+        border_in_loss = 0.5 * abs_border_diff * abs_border_diff * border_sign + \
+                         (abs_border_diff - 0.5) * (1.0 - border_sign)
+        border_out_loss = l_border_norm_split * border_in_loss
+        border_loss = paddle.sum(border_out_loss * l_border_score * l_border_mask) / \
+                      (paddle.sum(l_border_score * l_border_mask) + 1e-5)
+        return border_loss
+
+    def direction_loss(self, f_direction, l_direction, l_score, l_mask):
+        l_direction_split, l_direction_norm = paddle.tensor.split(
+            l_direction, num_or_sections=[2, 1], axis=1)
+        f_direction_split = f_direction
+        b, c, h, w = l_direction_norm.shape
+        l_direction_norm_split = paddle.expand(
+            x=l_direction_norm, shape=[b, 2 * c, h, w])
+        b, c, h, w = l_score.shape
+        l_direction_score = paddle.expand(x=l_score, shape=[b, 2 * c, h, w])
+        b, c, h, w = l_mask.shape
+        l_direction_mask = paddle.expand(x=l_mask, shape=[b, 2 * c, h, w])
+        direction_diff = l_direction_split - f_direction_split
+        abs_direction_diff = paddle.abs(direction_diff)
+        direction_sign = abs_direction_diff < 1.0
+        direction_sign = paddle.cast(direction_sign, dtype='float32')
+        direction_sign.stop_gradient = True
+        direction_in_loss = 0.5 * abs_direction_diff * abs_direction_diff * direction_sign + \
+                            (abs_direction_diff - 0.5) * (1.0 - direction_sign)
+        direction_out_loss = l_direction_norm_split * direction_in_loss
+        direction_loss = paddle.sum(direction_out_loss * l_direction_score * l_direction_mask) / \
+                         (paddle.sum(l_direction_score * l_direction_mask) + 1e-5)
+        return direction_loss
+
+    def ctcloss(self, f_char, tcl_pos, tcl_mask, tcl_label, label_t):
+        f_char = paddle.transpose(f_char, [0, 2, 3, 1])
+        tcl_pos = paddle.reshape(tcl_pos, [-1, 3])
+        tcl_pos = paddle.cast(tcl_pos, dtype=int)
+        f_tcl_char = paddle.gather_nd(f_char, tcl_pos)
+        f_tcl_char = paddle.reshape(f_tcl_char,
+                                    [-1, 64, 37])  # len(Lexicon_Table)+1
+        f_tcl_char_fg, f_tcl_char_bg = paddle.split(f_tcl_char, [36, 1], axis=2)
+        f_tcl_char_bg = f_tcl_char_bg * tcl_mask + (1.0 - tcl_mask) * 20.0
+        b, c, l = tcl_mask.shape
+        tcl_mask_fg = paddle.expand(x=tcl_mask, shape=[b, c, 36 * l])
+        tcl_mask_fg.stop_gradient = True
+        f_tcl_char_fg = f_tcl_char_fg * tcl_mask_fg + (1.0 - tcl_mask_fg) * (
+            -20.0)
+        f_tcl_char_mask = paddle.concat([f_tcl_char_fg, f_tcl_char_bg], axis=2)
+        f_tcl_char_ld = paddle.transpose(f_tcl_char_mask, (1, 0, 2))
+        N, B, _ = f_tcl_char_ld.shape
+        input_lengths = paddle.to_tensor([N] * B, dtype='int64')
+        cost = paddle.nn.functional.ctc_loss(
+            log_probs=f_tcl_char_ld,
+            labels=tcl_label,
+            input_lengths=input_lengths,
+            label_lengths=label_t,
+            blank=self.pad_num,
+            reduction='none')
+        cost = cost.mean()
+        return cost
+
+    def forward(self, predicts, labels):
+        images, tcl_maps, tcl_label_maps, border_maps \
+            , direction_maps, training_masks, label_list, pos_list, pos_mask = labels
+        # for all the batch_size
+        pos_list, pos_mask, label_list, label_t = pre_process(
+            label_list, pos_list, pos_mask, self.max_text_length,
+            self.max_text_nums, self.pad_num, self.tcl_bs)
+
+        f_score, f_border, f_direction, f_char = predicts['f_score'], predicts['f_border'], predicts['f_direction'], \
+                                                 predicts['f_char']
+        score_loss = self.dice_loss(f_score, tcl_maps, training_masks)
+        border_loss = self.border_loss(f_border, border_maps, tcl_maps,
+                                       training_masks)
+        direction_loss = self.direction_loss(f_direction, direction_maps,
+                                             tcl_maps, training_masks)
+        ctc_loss = self.ctcloss(f_char, pos_list, pos_mask, label_list, label_t)
+        loss_all = score_loss + border_loss + direction_loss + 5 * ctc_loss
+
+        losses = {
+            'loss': loss_all,
+            "score_loss": score_loss,
+            "border_loss": border_loss,
+            "direction_loss": direction_loss,
+            "ctc_loss": ctc_loss
+        }
+        return losses

ppocr/metrics/__init__.py

@@ -26,8 +26,9 @@ def build_metric(config):
     from .det_metric import DetMetric
     from .rec_metric import RecMetric
     from .cls_metric import ClsMetric
+    from .e2e_metric import E2EMetric
 
-    support_dict = ['DetMetric', 'RecMetric', 'ClsMetric']
+    support_dict = ['DetMetric', 'RecMetric', 'ClsMetric', 'E2EMetric']
 
     config = copy.deepcopy(config)
     module_name = config.pop('name')

ppocr/metrics/e2e_metric.py

+# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from __future__ import absolute_import
+from __future__ import division
+from __future__ import print_function
+
+__all__ = ['E2EMetric']
+
+from ppocr.utils.e2e_metric.Deteval import get_socre, combine_results
+from ppocr.utils.e2e_utils.extract_textpoint import get_dict
+
+
+class E2EMetric(object):
+    def __init__(self,
+                 character_dict_path,
+                 main_indicator='f_score_e2e',
+                 **kwargs):
+        self.label_list = get_dict(character_dict_path)
+        self.max_index = len(self.label_list)
+        self.main_indicator = main_indicator
+        self.reset()
+
+    def __call__(self, preds, batch, **kwargs):
+        temp_gt_polyons_batch = batch[2]
+        temp_gt_strs_batch = batch[3]
+        ignore_tags_batch = batch[4]
+        gt_polyons_batch = []
+        gt_strs_batch = []
+
+        temp_gt_polyons_batch = temp_gt_polyons_batch[0].tolist()
+        for temp_list in temp_gt_polyons_batch:
+            t = []
+            for index in temp_list:
+                if index[0] != -1 and index[1] != -1:
+                    t.append(index)
+            gt_polyons_batch.append(t)
+
+        temp_gt_strs_batch = temp_gt_strs_batch[0].tolist()
+        for temp_list in temp_gt_strs_batch:
+            t = ""
+            for index in temp_list:
+                if index < self.max_index:
+                    t += self.label_list[index]
+            gt_strs_batch.append(t)
+
+        for pred, gt_polyons, gt_strs, ignore_tags in zip(
+            [preds], [gt_polyons_batch], [gt_strs_batch], ignore_tags_batch):
+            # prepare gt
+            gt_info_list = [{
+                'points': gt_polyon,
+                'text': gt_str,
+                'ignore': ignore_tag
+            } for gt_polyon, gt_str, ignore_tag in
+                            zip(gt_polyons, gt_strs, ignore_tags)]
+            # prepare det
+            e2e_info_list = [{
+                'points': det_polyon,
+                'text': pred_str
+            } for det_polyon, pred_str in zip(pred['points'], pred['strs'])]
+            result = get_socre(gt_info_list, e2e_info_list)
+            self.results.append(result)
+
+    def get_metric(self):
+        metircs = combine_results(self.results)
+        self.reset()
+        return metircs
+
+    def reset(self):
+        self.results = []  # clear results

ppocr/metrics/eval_det_iou.py

@@ -150,7 +150,7 @@ class DetectionIoUEvaluator(object):
                             pairs.append({'gt': gtNum, 'det': detNum})
                             detMatchedNums.append(detNum)
                             evaluationLog += "Match GT #" + \
-                                str(gtNum) + " with Det #" + str(detNum) + "\n"
+                                             str(gtNum) + " with Det #" + str(detNum) + "\n"
 
         numGtCare = (len(gtPols) - len(gtDontCarePolsNum))
         numDetCare = (len(detPols) - len(detDontCarePolsNum))
@@ -162,7 +162,7 @@ class DetectionIoUEvaluator(object):
             precision = 0 if numDetCare == 0 else float(detMatched) / numDetCare
 
         hmean = 0 if (precision + recall) == 0 else 2.0 * \
-            precision * recall / (precision + recall)
+                                                    precision * recall / (precision + recall)
 
         matchedSum += detMatched
         numGlobalCareGt += numGtCare
@@ -200,7 +200,8 @@ class DetectionIoUEvaluator(object):
         methodPrecision = 0 if numGlobalCareDet == 0 else float(
             matchedSum) / numGlobalCareDet
         methodHmean = 0 if methodRecall + methodPrecision == 0 else 2 * \
-            methodRecall * methodPrecision / (methodRecall + methodPrecision)
+                                                                    methodRecall * methodPrecision / (
+                                                                            methodRecall + methodPrecision)
         # print(methodRecall, methodPrecision, methodHmean)
         # sys.exit(-1)
         methodMetrics = {

ppocr/modeling/backbones/__init__.py

@@ -26,6 +26,9 @@ def build_backbone(config, model_type):
         from .rec_resnet_vd import ResNet
         from .rec_resnet_fpn import ResNetFPN
         support_dict = ['MobileNetV3', 'ResNet', 'ResNetFPN']
+    elif model_type == 'e2e':
+        from .e2e_resnet_vd_pg import ResNet
+        support_dict = ['ResNet']
     else:
         raise NotImplementedError
 

ppocr/modeling/backbones/e2e_resnet_vd_pg.py

+# copyright (c) 2021 PaddlePaddle Authors. All Rights Reserve.
+#
+# 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.
+
+from __future__ import absolute_import
+from __future__ import division
+from __future__ import print_function
+
+import paddle
+from paddle import ParamAttr
+import paddle.nn as nn
+import paddle.nn.functional as F
+
+__all__ = ["ResNet"]
+
+
+class ConvBNLayer(nn.Layer):
+    def __init__(
+            self,
+            in_channels,
+            out_channels,
+            kernel_size,
+            stride=1,
+            groups=1,
+            is_vd_mode=False,
+            act=None,
+            name=None, ):
+        super(ConvBNLayer, self).__init__()
+
+        self.is_vd_mode = is_vd_mode
+        self._pool2d_avg = nn.AvgPool2D(
+            kernel_size=2, stride=2, padding=0, ceil_mode=True)
+        self._conv = nn.Conv2D(
+            in_channels=in_channels,
+            out_channels=out_channels,
+            kernel_size=kernel_size,
+            stride=stride,
+            padding=(kernel_size - 1) // 2,
+            groups=groups,
+            weight_attr=ParamAttr(name=name + "_weights"),
+            bias_attr=False)
+        if name == "conv1":
+            bn_name = "bn_" + name
+        else:
+            bn_name = "bn" + name[3:]
+        self._batch_norm = nn.BatchNorm(
+            out_channels,
+            act=act,
+            param_attr=ParamAttr(name=bn_name + '_scale'),
+            bias_attr=ParamAttr(bn_name + '_offset'),
+            moving_mean_name=bn_name + '_mean',
+            moving_variance_name=bn_name + '_variance')
+
+    def forward(self, inputs):
+        y = self._conv(inputs)
+        y = self._batch_norm(y)
+        return y
+
+
+class BottleneckBlock(nn.Layer):
+    def __init__(self,
+                 in_channels,
+                 out_channels,
+                 stride,
+                 shortcut=True,
+                 if_first=False,
+                 name=None):
+        super(BottleneckBlock, self).__init__()
+
+        self.conv0 = ConvBNLayer(
+            in_channels=in_channels,
+            out_channels=out_channels,
+            kernel_size=1,
+            act='relu',
+            name=name + "_branch2a")
+        self.conv1 = ConvBNLayer(
+            in_channels=out_channels,
+            out_channels=out_channels,
+            kernel_size=3,
+            stride=stride,
+            act='relu',
+            name=name + "_branch2b")
+        self.conv2 = ConvBNLayer(
+            in_channels=out_channels,
+            out_channels=out_channels * 4,
+            kernel_size=1,
+            act=None,
+            name=name + "_branch2c")
+
+        if not shortcut:
+            self.short = ConvBNLayer(
+                in_channels=in_channels,
+                out_channels=out_channels * 4,
+                kernel_size=1,
+                stride=stride,
+                is_vd_mode=False if if_first else True,
+                name=name + "_branch1")
+
+        self.shortcut = shortcut
+
+    def forward(self, inputs):
+        y = self.conv0(inputs)
+        conv1 = self.conv1(y)
+        conv2 = self.conv2(conv1)
+
+        if self.shortcut:
+            short = inputs
+        else:
+            short = self.short(inputs)
+        y = paddle.add(x=short, y=conv2)
+        y = F.relu(y)
+        return y
+
+
+class BasicBlock(nn.Layer):
+    def __init__(self,
+                 in_channels,
+                 out_channels,
+                 stride,
+                 shortcut=True,
+                 if_first=False,
+                 name=None):
+        super(BasicBlock, self).__init__()
+        self.stride = stride
+        self.conv0 = ConvBNLayer(
+            in_channels=in_channels,
+            out_channels=out_channels,
+            kernel_size=3,
+            stride=stride,
+            act='relu',
+            name=name + "_branch2a")
+        self.conv1 = ConvBNLayer(
+            in_channels=out_channels,
+            out_channels=out_channels,
+            kernel_size=3,
+            act=None,
+            name=name + "_branch2b")
+
+        if not shortcut:
+            self.short = ConvBNLayer(
+                in_channels=in_channels,
+                out_channels=out_channels,
+                kernel_size=1,
+                stride=1,
+                is_vd_mode=False if if_first else True,
+                name=name + "_branch1")
+
+        self.shortcut = shortcut
+
+    def forward(self, inputs):
+        y = self.conv0(inputs)
+        conv1 = self.conv1(y)
+
+        if self.shortcut:
+            short = inputs
+        else:
+            short = self.short(inputs)
+        y = paddle.add(x=short, y=conv1)
+        y = F.relu(y)
+        return y
+
+
+class ResNet(nn.Layer):
+    def __init__(self, in_channels=3, layers=50, **kwargs):
+        super(ResNet, self).__init__()
+
+        self.layers = layers
+        supported_layers = [18, 34, 50, 101, 152, 200]
+        assert layers in supported_layers, \
+            "supported layers are {} but input layer is {}".format(
+                supported_layers, layers)
+
+        if layers == 18:
+            depth = [2, 2, 2, 2]
+        elif layers == 34 or layers == 50:
+            # depth = [3, 4, 6, 3]
+            depth = [3, 4, 6, 3, 3]
+        elif layers == 101:
+            depth = [3, 4, 23, 3]
+        elif layers == 152:
+            depth = [3, 8, 36, 3]
+        elif layers == 200:
+            depth = [3, 12, 48, 3]
+        num_channels = [64, 256, 512, 1024,
+                        2048] if layers >= 50 else [64, 64, 128, 256]
+        num_filters = [64, 128, 256, 512, 512]
+
+        self.conv1_1 = ConvBNLayer(
+            in_channels=in_channels,
+            out_channels=64,
+            kernel_size=7,
+            stride=2,
+            act='relu',
+            name="conv1_1")
+        self.pool2d_max = nn.MaxPool2D(kernel_size=3, stride=2, padding=1)
+
+        self.stages = []
+        self.out_channels = [3, 64]
+        # num_filters = [64, 128, 256, 512, 512]
+        if layers >= 50:
+            for block in range(len(depth)):
+                block_list = []
+                shortcut = False
+                for i in range(depth[block]):
+                    if layers in [101, 152] and block == 2:
+                        if i == 0:
+                            conv_name = "res" + str(block + 2) + "a"
+                        else:
+                            conv_name = "res" + str(block + 2) + "b" + str(i)
+                    else:
+                        conv_name = "res" + str(block + 2) + chr(97 + i)
+                    bottleneck_block = self.add_sublayer(
+                        'bb_%d_%d' % (block, i),
+                        BottleneckBlock(
+                            in_channels=num_channels[block]
+                            if i == 0 else num_filters[block] * 4,
+                            out_channels=num_filters[block],
+                            stride=2 if i == 0 and block != 0 else 1,
+                            shortcut=shortcut,
+                            if_first=block == i == 0,
+                            name=conv_name))
+                    shortcut = True
+                    block_list.append(bottleneck_block)
+                self.out_channels.append(num_filters[block] * 4)
+                self.stages.append(nn.Sequential(*block_list))
+        else:
+            for block in range(len(depth)):
+                block_list = []
+                shortcut = False
+                for i in range(depth[block]):
+                    conv_name = "res" + str(block + 2) + chr(97 + i)
+                    basic_block = self.add_sublayer(
+                        'bb_%d_%d' % (block, i),
+                        BasicBlock(
+                            in_channels=num_channels[block]
+                            if i == 0 else num_filters[block],
+                            out_channels=num_filters[block],
+                            stride=2 if i == 0 and block != 0 else 1,
+                            shortcut=shortcut,
+                            if_first=block == i == 0,
+                            name=conv_name))
+                    shortcut = True
+                    block_list.append(basic_block)
+                self.out_channels.append(num_filters[block])
+                self.stages.append(nn.Sequential(*block_list))
+
+    def forward(self, inputs):
+        out = [inputs]
+        y = self.conv1_1(inputs)
+        out.append(y)
+        y = self.pool2d_max(y)
+        for block in self.stages:
+            y = block(y)
+            out.append(y)
+        return out

ppocr/modeling/heads/__init__.py

@@ -20,6 +20,7 @@ def build_head(config):
     from .det_db_head import DBHead
     from .det_east_head import EASTHead
     from .det_sast_head import SASTHead
+    from .e2e_pg_head import PGHead
 
     # rec head
     from .rec_ctc_head import CTCHead
@@ -30,8 +31,8 @@ def build_head(config):
     from .cls_head import ClsHead
     support_dict = [
         'DBHead', 'EASTHead', 'SASTHead', 'CTCHead', 'ClsHead', 'AttentionHead',
-        'SRNHead'
-    ]
+        'SRNHead', 'PGHead']
+
 
     module_name = config.pop('name')
     assert module_name in support_dict, Exception('head only support {}'.format(

ppocr/modeling/heads/e2e_pg_head.py

+# copyright (c) 2021 PaddlePaddle Authors. All Rights Reserve.
+#
+# 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.
+
+from __future__ import absolute_import
+from __future__ import division
+from __future__ import print_function
+
+import math
+import paddle
+from paddle import nn
+import paddle.nn.functional as F
+from paddle import ParamAttr
+
+
+class ConvBNLayer(nn.Layer):
+    def __init__(self,
+                 in_channels,
+                 out_channels,
+                 kernel_size,
+                 stride,
+                 padding,
+                 groups=1,
+                 if_act=True,
+                 act=None,
+                 name=None):
+        super(ConvBNLayer, self).__init__()
+        self.if_act = if_act
+        self.act = act
+        self.conv = nn.Conv2D(
+            in_channels=in_channels,
+            out_channels=out_channels,
+            kernel_size=kernel_size,
+            stride=stride,
+            padding=padding,
+            groups=groups,
+            weight_attr=ParamAttr(name=name + '_weights'),
+            bias_attr=False)
+
+        self.bn = nn.BatchNorm(
+            num_channels=out_channels,
+            act=act,
+            param_attr=ParamAttr(name="bn_" + name + "_scale"),
+            bias_attr=ParamAttr(name="bn_" + name + "_offset"),
+            moving_mean_name="bn_" + name + "_mean",
+            moving_variance_name="bn_" + name + "_variance",
+            use_global_stats=False)
+
+    def forward(self, x):
+        x = self.conv(x)
+        x = self.bn(x)
+        return x
+
+
+class PGHead(nn.Layer):
+    """
+    """
+
+    def __init__(self, in_channels, **kwargs):
+        super(PGHead, self).__init__()
+        self.conv_f_score1 = ConvBNLayer(
+            in_channels=in_channels,
+            out_channels=64,
+            kernel_size=1,
+            stride=1,
+            padding=0,
+            act='relu',
+            name="conv_f_score{}".format(1))
+        self.conv_f_score2 = ConvBNLayer(
+            in_channels=64,
+            out_channels=64,
+            kernel_size=3,
+            stride=1,
+            padding=1,
+            act='relu',
+            name="conv_f_score{}".format(2))
+        self.conv_f_score3 = ConvBNLayer(
+            in_channels=64,
+            out_channels=128,
+            kernel_size=1,
+            stride=1,
+            padding=0,
+            act='relu',
+            name="conv_f_score{}".format(3))
+
+        self.conv1 = nn.Conv2D(
+            in_channels=128,
+            out_channels=1,
+            kernel_size=3,
+            stride=1,
+            padding=1,
+            groups=1,
+            weight_attr=ParamAttr(name="conv_f_score{}".format(4)),
+            bias_attr=False)
+
+        self.conv_f_boder1 = ConvBNLayer(
+            in_channels=in_channels,
+            out_channels=64,
+            kernel_size=1,
+            stride=1,
+            padding=0,
+            act='relu',
+            name="conv_f_boder{}".format(1))
+        self.conv_f_boder2 = ConvBNLayer(
+            in_channels=64,
+            out_channels=64,
+            kernel_size=3,
+            stride=1,
+            padding=1,
+            act='relu',
+            name="conv_f_boder{}".format(2))
+        self.conv_f_boder3 = ConvBNLayer(
+            in_channels=64,
+            out_channels=128,
+            kernel_size=1,
+            stride=1,
+            padding=0,
+            act='relu',
+            name="conv_f_boder{}".format(3))
+        self.conv2 = nn.Conv2D(
+            in_channels=128,
+            out_channels=4,
+            kernel_size=3,
+            stride=1,
+            padding=1,
+            groups=1,
+            weight_attr=ParamAttr(name="conv_f_boder{}".format(4)),
+            bias_attr=False)
+        self.conv_f_char1 = ConvBNLayer(
+            in_channels=in_channels,
+            out_channels=128,
+            kernel_size=1,
+            stride=1,
+            padding=0,
+            act='relu',
+            name="conv_f_char{}".format(1))
+        self.conv_f_char2 = ConvBNLayer(
+            in_channels=128,
+            out_channels=128,
+            kernel_size=3,
+            stride=1,
+            padding=1,
+            act='relu',
+            name="conv_f_char{}".format(2))
+        self.conv_f_char3 = ConvBNLayer(
+            in_channels=128,
+            out_channels=256,
+            kernel_size=1,
+            stride=1,
+            padding=0,
+            act='relu',
+            name="conv_f_char{}".format(3))
+        self.conv_f_char4 = ConvBNLayer(
+            in_channels=256,
+            out_channels=256,
+            kernel_size=3,
+            stride=1,
+            padding=1,
+            act='relu',
+            name="conv_f_char{}".format(4))
+        self.conv_f_char5 = ConvBNLayer(
+            in_channels=256,
+            out_channels=256,
+            kernel_size=1,
+            stride=1,
+            padding=0,
+            act='relu',
+            name="conv_f_char{}".format(5))
+        self.conv3 = nn.Conv2D(
+            in_channels=256,
+            out_channels=37,
+            kernel_size=3,
+            stride=1,
+            padding=1,
+            groups=1,
+            weight_attr=ParamAttr(name="conv_f_char{}".format(6)),
+            bias_attr=False)
+
+        self.conv_f_direc1 = ConvBNLayer(
+            in_channels=in_channels,
+            out_channels=64,
+            kernel_size=1,
+            stride=1,
+            padding=0,
+            act='relu',
+            name="conv_f_direc{}".format(1))
+        self.conv_f_direc2 = ConvBNLayer(
+            in_channels=64,
+            out_channels=64,
+            kernel_size=3,
+            stride=1,
+            padding=1,
+            act='relu',
+            name="conv_f_direc{}".format(2))
+        self.conv_f_direc3 = ConvBNLayer(
+            in_channels=64,
+            out_channels=128,
+            kernel_size=1,
+            stride=1,
+            padding=0,
+            act='relu',
+            name="conv_f_direc{}".format(3))
+        self.conv4 = nn.Conv2D(
+            in_channels=128,
+            out_channels=2,
+            kernel_size=3,
+            stride=1,
+            padding=1,
+            groups=1,
+            weight_attr=ParamAttr(name="conv_f_direc{}".format(4)),
+            bias_attr=False)
+
+    def forward(self, x):
+        f_score = self.conv_f_score1(x)
+        f_score = self.conv_f_score2(f_score)
+        f_score = self.conv_f_score3(f_score)
+        f_score = self.conv1(f_score)
+        f_score = F.sigmoid(f_score)
+
+        # f_border
+        f_border = self.conv_f_boder1(x)
+        f_border = self.conv_f_boder2(f_border)
+        f_border = self.conv_f_boder3(f_border)
+        f_border = self.conv2(f_border)
+
+        f_char = self.conv_f_char1(x)
+        f_char = self.conv_f_char2(f_char)
+        f_char = self.conv_f_char3(f_char)
+        f_char = self.conv_f_char4(f_char)
+        f_char = self.conv_f_char5(f_char)
+        f_char = self.conv3(f_char)
+
+        f_direction = self.conv_f_direc1(x)
+        f_direction = self.conv_f_direc2(f_direction)
+        f_direction = self.conv_f_direc3(f_direction)
+        f_direction = self.conv4(f_direction)
+
+        predicts = {}
+        predicts['f_score'] = f_score
+        predicts['f_border'] = f_border
+        predicts['f_char'] = f_char
+        predicts['f_direction'] = f_direction
+        return predicts

ppocr/modeling/necks/__init__.py

@@ -14,12 +14,14 @@
 
 __all__ = ['build_neck']
 
+
 def build_neck(config):
     from .db_fpn import DBFPN
     from .east_fpn import EASTFPN
     from .sast_fpn import SASTFPN
     from .rnn import SequenceEncoder
-    support_dict = ['DBFPN', 'EASTFPN', 'SASTFPN', 'SequenceEncoder']
+    from .pg_fpn import PGFPN
+    support_dict = ['DBFPN', 'EASTFPN', 'SASTFPN', 'SequenceEncoder', 'PGFPN']
 
     module_name = config.pop('name')
     assert module_name in support_dict, Exception('neck only support {}'.format(

ppocr/modeling/necks/pg_fpn.py

+# copyright (c) 2021 PaddlePaddle Authors. All Rights Reserve.
+#
+# 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.
+
+from __future__ import absolute_import
+from __future__ import division
+from __future__ import print_function
+
+import paddle
+from paddle import nn
+import paddle.nn.functional as F
+from paddle import ParamAttr
+
+
+class ConvBNLayer(nn.Layer):
+    def __init__(self,
+                 in_channels,
+                 out_channels,
+                 kernel_size,
+                 stride=1,
+                 groups=1,
+                 is_vd_mode=False,
+                 act=None,
+                 name=None):
+        super(ConvBNLayer, self).__init__()
+
+        self.is_vd_mode = is_vd_mode
+        self._pool2d_avg = nn.AvgPool2D(
+            kernel_size=2, stride=2, padding=0, ceil_mode=True)
+        self._conv = nn.Conv2D(
+            in_channels=in_channels,
+            out_channels=out_channels,
+            kernel_size=kernel_size,
+            stride=stride,
+            padding=(kernel_size - 1) // 2,
+            groups=groups,
+            weight_attr=ParamAttr(name=name + "_weights"),
+            bias_attr=False)
+        if name == "conv1":
+            bn_name = "bn_" + name
+        else:
+            bn_name = "bn" + name[3:]
+        self._batch_norm = nn.BatchNorm(
+            out_channels,
+            act=act,
+            param_attr=ParamAttr(name=bn_name + '_scale'),
+            bias_attr=ParamAttr(bn_name + '_offset'),
+            moving_mean_name=bn_name + '_mean',
+            moving_variance_name=bn_name + '_variance',
+            use_global_stats=False)
+
+    def forward(self, inputs):
+        y = self._conv(inputs)
+        y = self._batch_norm(y)
+        return y
+
+
+class DeConvBNLayer(nn.Layer):
+    def __init__(self,
+                 in_channels,
+                 out_channels,
+                 kernel_size=4,
+                 stride=2,
+                 padding=1,
+                 groups=1,
+                 if_act=True,
+                 act=None,
+                 name=None):
+        super(DeConvBNLayer, self).__init__()
+
+        self.if_act = if_act
+        self.act = act
+        self.deconv = nn.Conv2DTranspose(
+            in_channels=in_channels,
+            out_channels=out_channels,
+            kernel_size=kernel_size,
+            stride=stride,
+            padding=padding,
+            groups=groups,
+            weight_attr=ParamAttr(name=name + '_weights'),
+            bias_attr=False)
+        self.bn = nn.BatchNorm(
+            num_channels=out_channels,
+            act=act,
+            param_attr=ParamAttr(name="bn_" + name + "_scale"),
+            bias_attr=ParamAttr(name="bn_" + name + "_offset"),
+            moving_mean_name="bn_" + name + "_mean",
+            moving_variance_name="bn_" + name + "_variance",
+            use_global_stats=False)
+
+    def forward(self, x):
+        x = self.deconv(x)
+        x = self.bn(x)
+        return x
+
+
+class PGFPN(nn.Layer):
+    def __init__(self, in_channels, **kwargs):
+        super(PGFPN, self).__init__()
+        num_inputs = [2048, 2048, 1024, 512, 256]
+        num_outputs = [256, 256, 192, 192, 128]
+        self.out_channels = 128
+        self.conv_bn_layer_1 = ConvBNLayer(
+            in_channels=3,
+            out_channels=32,
+            kernel_size=3,
+            stride=1,
+            act=None,
+            name='FPN_d1')
+        self.conv_bn_layer_2 = ConvBNLayer(
+            in_channels=64,
+            out_channels=64,
+            kernel_size=3,
+            stride=1,
+            act=None,
+            name='FPN_d2')
+        self.conv_bn_layer_3 = ConvBNLayer(
+            in_channels=256,
+            out_channels=128,
+            kernel_size=3,
+            stride=1,
+            act=None,
+            name='FPN_d3')
+        self.conv_bn_layer_4 = ConvBNLayer(
+            in_channels=32,
+            out_channels=64,
+            kernel_size=3,
+            stride=2,
+            act=None,
+            name='FPN_d4')
+        self.conv_bn_layer_5 = ConvBNLayer(
+            in_channels=64,
+            out_channels=64,
+            kernel_size=3,
+            stride=1,
+            act='relu',
+            name='FPN_d5')
+        self.conv_bn_layer_6 = ConvBNLayer(
+            in_channels=64,
+            out_channels=128,
+            kernel_size=3,
+            stride=2,
+            act=None,
+            name='FPN_d6')
+        self.conv_bn_layer_7 = ConvBNLayer(
+            in_channels=128,
+            out_channels=128,
+            kernel_size=3,
+            stride=1,
+            act='relu',
+            name='FPN_d7')
+        self.conv_bn_layer_8 = ConvBNLayer(
+            in_channels=128,
+            out_channels=128,
+            kernel_size=1,
+            stride=1,
+            act=None,
+            name='FPN_d8')
+
+        self.conv_h0 = ConvBNLayer(
+            in_channels=num_inputs[0],
+            out_channels=num_outputs[0],
+            kernel_size=1,
+            stride=1,
+            act=None,
+            name="conv_h{}".format(0))
+        self.conv_h1 = ConvBNLayer(
+            in_channels=num_inputs[1],
+            out_channels=num_outputs[1],
+            kernel_size=1,
+            stride=1,
+            act=None,
+            name="conv_h{}".format(1))
+        self.conv_h2 = ConvBNLayer(
+            in_channels=num_inputs[2],
+            out_channels=num_outputs[2],
+            kernel_size=1,
+            stride=1,
+            act=None,
+            name="conv_h{}".format(2))
+        self.conv_h3 = ConvBNLayer(
+            in_channels=num_inputs[3],
+            out_channels=num_outputs[3],
+            kernel_size=1,
+            stride=1,
+            act=None,
+            name="conv_h{}".format(3))
+        self.conv_h4 = ConvBNLayer(
+            in_channels=num_inputs[4],
+            out_channels=num_outputs[4],
+            kernel_size=1,
+            stride=1,
+            act=None,
+            name="conv_h{}".format(4))
+
+        self.dconv0 = DeConvBNLayer(
+            in_channels=num_outputs[0],
+            out_channels=num_outputs[0 + 1],
+            name="dconv_{}".format(0))
+        self.dconv1 = DeConvBNLayer(
+            in_channels=num_outputs[1],
+            out_channels=num_outputs[1 + 1],
+            act=None,
+            name="dconv_{}".format(1))
+        self.dconv2 = DeConvBNLayer(
+            in_channels=num_outputs[2],
+            out_channels=num_outputs[2 + 1],
+            act=None,
+            name="dconv_{}".format(2))
+        self.dconv3 = DeConvBNLayer(
+            in_channels=num_outputs[3],
+            out_channels=num_outputs[3 + 1],
+            act=None,
+            name="dconv_{}".format(3))
+        self.conv_g1 = ConvBNLayer(
+            in_channels=num_outputs[1],
+            out_channels=num_outputs[1],
+            kernel_size=3,
+            stride=1,
+            act='relu',
+            name="conv_g{}".format(1))
+        self.conv_g2 = ConvBNLayer(
+            in_channels=num_outputs[2],
+            out_channels=num_outputs[2],
+            kernel_size=3,
+            stride=1,
+            act='relu',
+            name="conv_g{}".format(2))
+        self.conv_g3 = ConvBNLayer(
+            in_channels=num_outputs[3],
+            out_channels=num_outputs[3],
+            kernel_size=3,
+            stride=1,
+            act='relu',
+            name="conv_g{}".format(3))
+        self.conv_g4 = ConvBNLayer(
+            in_channels=num_outputs[4],
+            out_channels=num_outputs[4],
+            kernel_size=3,
+            stride=1,
+            act='relu',
+            name="conv_g{}".format(4))
+        self.convf = ConvBNLayer(
+            in_channels=num_outputs[4],
+            out_channels=num_outputs[4],
+            kernel_size=1,
+            stride=1,
+            act=None,
+            name="conv_f{}".format(4))
+
+    def forward(self, x):
+        c0, c1, c2, c3, c4, c5, c6 = x
+        # FPN_Down_Fusion
+        f = [c0, c1, c2]
+        g = [None, None, None]
+        h = [None, None, None]
+        h[0] = self.conv_bn_layer_1(f[0])
+        h[1] = self.conv_bn_layer_2(f[1])
+        h[2] = self.conv_bn_layer_3(f[2])
+
+        g[0] = self.conv_bn_layer_4(h[0])
+        g[1] = paddle.add(g[0], h[1])
+        g[1] = F.relu(g[1])
+        g[1] = self.conv_bn_layer_5(g[1])
+        g[1] = self.conv_bn_layer_6(g[1])
+
+        g[2] = paddle.add(g[1], h[2])
+        g[2] = F.relu(g[2])
+        g[2] = self.conv_bn_layer_7(g[2])
+        f_down = self.conv_bn_layer_8(g[2])
+
+        # FPN UP Fusion
+        f1 = [c6, c5, c4, c3, c2]
+        g = [None, None, None, None, None]
+        h = [None, None, None, None, None]
+        h[0] = self.conv_h0(f1[0])
+        h[1] = self.conv_h1(f1[1])
+        h[2] = self.conv_h2(f1[2])
+        h[3] = self.conv_h3(f1[3])
+        h[4] = self.conv_h4(f1[4])
+
+        g[0] = self.dconv0(h[0])
+        g[1] = paddle.add(g[0], h[1])
+        g[1] = F.relu(g[1])
+        g[1] = self.conv_g1(g[1])
+        g[1] = self.dconv1(g[1])
+
+        g[2] = paddle.add(g[1], h[2])
+        g[2] = F.relu(g[2])
+        g[2] = self.conv_g2(g[2])
+        g[2] = self.dconv2(g[2])
+
+        g[3] = paddle.add(g[2], h[3])
+        g[3] = F.relu(g[3])
+        g[3] = self.conv_g3(g[3])
+        g[3] = self.dconv3(g[3])
+
+        g[4] = paddle.add(x=g[3], y=h[4])
+        g[4] = F.relu(g[4])
+        g[4] = self.conv_g4(g[4])
+        f_up = self.convf(g[4])
+        f_common = paddle.add(f_down, f_up)
+        f_common = F.relu(f_common)
+        return f_common

ppocr/postprocess/__init__.py

@@ -28,10 +28,11 @@ def build_post_process(config, global_config=None):
     from .sast_postprocess import SASTPostProcess
     from .rec_postprocess import CTCLabelDecode, AttnLabelDecode, SRNLabelDecode
     from .cls_postprocess import ClsPostProcess
+    from .pg_postprocess import PGPostProcess
 
     support_dict = [
         'DBPostProcess', 'EASTPostProcess', 'SASTPostProcess', 'CTCLabelDecode',
-        'AttnLabelDecode', 'ClsPostProcess', 'SRNLabelDecode'
+        'AttnLabelDecode', 'ClsPostProcess', 'SRNLabelDecode', 'PGPostProcess'
     ]
 
     config = copy.deepcopy(config)

ppocr/postprocess/pg_postprocess.py

+# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from __future__ import absolute_import
+from __future__ import division
+from __future__ import print_function
+
+import os
+import sys
+
+__dir__ = os.path.dirname(__file__)
+sys.path.append(__dir__)
+sys.path.append(os.path.join(__dir__, '..'))
+
+from ppocr.utils.e2e_utils.extract_textpoint import *
+from ppocr.utils.e2e_utils.visual import *
+import paddle
+
+
+class PGPostProcess(object):
+    """
+    The post process for PGNet.
+    """
+
+    def __init__(self, character_dict_path, valid_set, score_thresh, **kwargs):
+        self.Lexicon_Table = get_dict(character_dict_path)
+        self.valid_set = valid_set
+        self.score_thresh = score_thresh
+
+        # c++ la-nms is faster, but only support python 3.5
+        self.is_python35 = False
+        if sys.version_info.major == 3 and sys.version_info.minor == 5:
+            self.is_python35 = True
+
+    def __call__(self, outs_dict, shape_list):
+        p_score = outs_dict['f_score']
+        p_border = outs_dict['f_border']
+        p_char = outs_dict['f_char']
+        p_direction = outs_dict['f_direction']
+        if isinstance(p_score, paddle.Tensor):
+            p_score = p_score[0].numpy()
+            p_border = p_border[0].numpy()
+            p_direction = p_direction[0].numpy()
+            p_char = p_char[0].numpy()
+        else:
+            p_score = p_score[0]
+            p_border = p_border[0]
+            p_direction = p_direction[0]
+            p_char = p_char[0]
+        src_h, src_w, ratio_h, ratio_w = shape_list[0]
+        is_curved = self.valid_set == "totaltext"
+        instance_yxs_list = generate_pivot_list(
+            p_score,
+            p_char,
+            p_direction,
+            score_thresh=self.score_thresh,
+            is_backbone=True,
+            is_curved=is_curved)
+        p_char = paddle.to_tensor(np.expand_dims(p_char, axis=0))
+        char_seq_idx_set = []
+        for i in range(len(instance_yxs_list)):
+            gather_info_lod = paddle.to_tensor(instance_yxs_list[i])
+            f_char_map = paddle.transpose(p_char, [0, 2, 3, 1])
+            feature_seq = paddle.gather_nd(f_char_map, gather_info_lod)
+            feature_seq = np.expand_dims(feature_seq.numpy(), axis=0)
+            feature_len = [len(feature_seq[0])]
+            featyre_seq = paddle.to_tensor(feature_seq)
+            feature_len = np.array([feature_len]).astype(np.int64)
+            length = paddle.to_tensor(feature_len)
+            seq_pred = paddle.fluid.layers.ctc_greedy_decoder(
+                input=featyre_seq, blank=36, input_length=length)
+            seq_pred_str = seq_pred[0].numpy().tolist()[0]
+            seq_len = seq_pred[1].numpy()[0][0]
+            temp_t = []
+            for c in seq_pred_str[:seq_len]:
+                temp_t.append(c)
+            char_seq_idx_set.append(temp_t)
+        seq_strs = []
+        for char_idx_set in char_seq_idx_set:
+            pr_str = ''.join([self.Lexicon_Table[pos] for pos in char_idx_set])
+            seq_strs.append(pr_str)
+        poly_list = []
+        keep_str_list = []
+        all_point_list = []
+        all_point_pair_list = []
+        for yx_center_line, keep_str in zip(instance_yxs_list, seq_strs):
+            if len(yx_center_line) == 1:
+                yx_center_line.append(yx_center_line[-1])
+
+            offset_expand = 1.0
+            if self.valid_set == 'totaltext':
+                offset_expand = 1.2
+
+            point_pair_list = []
+            for batch_id, y, x in yx_center_line:
+                offset = p_border[:, y, x].reshape(2, 2)
+                if offset_expand != 1.0:
+                    offset_length = np.linalg.norm(
+                        offset, axis=1, keepdims=True)
+                    expand_length = np.clip(
+                        offset_length * (offset_expand - 1),
+                        a_min=0.5,
+                        a_max=3.0)
+                    offset_detal = offset / offset_length * expand_length
+                    offset = offset + offset_detal
+                ori_yx = np.array([y, x], dtype=np.float32)
+                point_pair = (ori_yx + offset)[:, ::-1] * 4.0 / np.array(
+                    [ratio_w, ratio_h]).reshape(-1, 2)
+                point_pair_list.append(point_pair)
+
+                all_point_list.append([
+                    int(round(x * 4.0 / ratio_w)),
+                    int(round(y * 4.0 / ratio_h))
+                ])
+                all_point_pair_list.append(point_pair.round().astype(np.int32)
+                                           .tolist())
+
+            detected_poly, pair_length_info = point_pair2poly(point_pair_list)
+            detected_poly = expand_poly_along_width(
+                detected_poly, shrink_ratio_of_width=0.2)
+            detected_poly[:, 0] = np.clip(
+                detected_poly[:, 0], a_min=0, a_max=src_w)
+            detected_poly[:, 1] = np.clip(
+                detected_poly[:, 1], a_min=0, a_max=src_h)
+
+            if len(keep_str) < 2:
+                continue
+
+            keep_str_list.append(keep_str)
+            if self.valid_set == 'partvgg':
+                middle_point = len(detected_poly) // 2
+                detected_poly = detected_poly[
+                    [0, middle_point - 1, middle_point, -1], :]
+                poly_list.append(detected_poly)
+            elif self.valid_set == 'totaltext':
+                poly_list.append(detected_poly)
+            else:
+                print('--> Not supported format.')
+                exit(-1)
+        data = {
+            'points': poly_list,
+            'strs': keep_str_list,
+        }
+        return data

ppocr/postprocess/sast_postprocess.py

@@ -18,6 +18,7 @@ from __future__ import print_function
 
 import os
 import sys
+
 __dir__ = os.path.dirname(__file__)
 sys.path.append(__dir__)
 sys.path.append(os.path.join(__dir__, '..'))
@@ -49,12 +50,12 @@ class SASTPostProcess(object):
         self.shrink_ratio_of_width = shrink_ratio_of_width
         self.expand_scale = expand_scale
         self.tcl_map_thresh = tcl_map_thresh
-        
+
         # c++ la-nms is faster, but only support python 3.5
         self.is_python35 = False
         if sys.version_info.major == 3 and sys.version_info.minor == 5:
             self.is_python35 = True
-            
+
     def point_pair2poly(self, point_pair_list):
         """
         Transfer vertical point_pairs into poly point in clockwise.
@@ -66,31 +67,42 @@ class SASTPostProcess(object):
             point_list[idx] = point_pair[0]
             point_list[point_num - 1 - idx] = point_pair[1]
         return np.array(point_list).reshape(-1, 2)
-    
-    def shrink_quad_along_width(self, quad, begin_width_ratio=0., end_width_ratio=1.):
+
+    def shrink_quad_along_width(self,
+                                quad,
+                                begin_width_ratio=0.,
+                                end_width_ratio=1.):
         """ 
         Generate shrink_quad_along_width.
         """
-        ratio_pair = np.array([[begin_width_ratio], [end_width_ratio]], dtype=np.float32)
+        ratio_pair = np.array(
+            [[begin_width_ratio], [end_width_ratio]], dtype=np.float32)
         p0_1 = quad[0] + (quad[1] - quad[0]) * ratio_pair
         p3_2 = quad[3] + (quad[2] - quad[3]) * ratio_pair
         return np.array([p0_1[0], p0_1[1], p3_2[1], p3_2[0]])
-    
+
     def expand_poly_along_width(self, poly, shrink_ratio_of_width=0.3):
         """
         expand poly along width.
         """
         point_num = poly.shape[0]
-        left_quad = np.array([poly[0], poly[1], poly[-2], poly[-1]], dtype=np.float32)
+        left_quad = np.array(
+            [poly[0], poly[1], poly[-2], poly[-1]], dtype=np.float32)
         left_ratio = -shrink_ratio_of_width * np.linalg.norm(left_quad[0] - left_quad[3]) / \
-                    (np.linalg.norm(left_quad[0] - left_quad[1]) + 1e-6)
-        left_quad_expand = self.shrink_quad_along_width(left_quad, left_ratio, 1.0)
-        right_quad = np.array([poly[point_num // 2 - 2], poly[point_num // 2 - 1],
-                            poly[point_num // 2], poly[point_num // 2 + 1]], dtype=np.float32)
+                     (np.linalg.norm(left_quad[0] - left_quad[1]) + 1e-6)
+        left_quad_expand = self.shrink_quad_along_width(left_quad, left_ratio,
+                                                        1.0)
+        right_quad = np.array(
+            [
+                poly[point_num // 2 - 2], poly[point_num // 2 - 1],
+                poly[point_num // 2], poly[point_num // 2 + 1]
+            ],
+            dtype=np.float32)
         right_ratio = 1.0 + \
-                    shrink_ratio_of_width * np.linalg.norm(right_quad[0] - right_quad[3]) / \
-                    (np.linalg.norm(right_quad[0] - right_quad[1]) + 1e-6)
-        right_quad_expand = self.shrink_quad_along_width(right_quad, 0.0, right_ratio)
+                      shrink_ratio_of_width * np.linalg.norm(right_quad[0] - right_quad[3]) / \
+                      (np.linalg.norm(right_quad[0] - right_quad[1]) + 1e-6)
+        right_quad_expand = self.shrink_quad_along_width(right_quad, 0.0,
+                                                         right_ratio)
         poly[0] = left_quad_expand[0]
         poly[-1] = left_quad_expand[-1]
         poly[point_num // 2 - 1] = right_quad_expand[1]
@@ -100,7 +112,7 @@ class SASTPostProcess(object):
     def restore_quad(self, tcl_map, tcl_map_thresh, tvo_map):
         """Restore quad."""
         xy_text = np.argwhere(tcl_map[:, :, 0] > tcl_map_thresh)
-        xy_text = xy_text[:, ::-1] # (n, 2)
+        xy_text = xy_text[:, ::-1]  # (n, 2)
 
         # Sort the text boxes via the y axis
         xy_text = xy_text[np.argsort(xy_text[:, 1])]
@@ -112,7 +124,7 @@ class SASTPostProcess(object):
         point_num = int(tvo_map.shape[-1] / 2)
         assert point_num == 4
         tvo_map = tvo_map[xy_text[:, 1], xy_text[:, 0], :]
-        xy_text_tile = np.tile(xy_text, (1, point_num)) # (n, point_num * 2)
+        xy_text_tile = np.tile(xy_text, (1, point_num))  # (n, point_num * 2)
         quads = xy_text_tile - tvo_map
 
         return scores, quads, xy_text
@@ -121,14 +133,12 @@ class SASTPostProcess(object):
         """
         compute area of a quad.
         """
-        edge = [
-            (quad[1][0] - quad[0][0]) * (quad[1][1] + quad[0][1]),
-            (quad[2][0] - quad[1][0]) * (quad[2][1] + quad[1][1]),
-            (quad[3][0] - quad[2][0]) * (quad[3][1] + quad[2][1]),
-            (quad[0][0] - quad[3][0]) * (quad[0][1] + quad[3][1])
-        ]
+        edge = [(quad[1][0] - quad[0][0]) * (quad[1][1] + quad[0][1]),
+                (quad[2][0] - quad[1][0]) * (quad[2][1] + quad[1][1]),
+                (quad[3][0] - quad[2][0]) * (quad[3][1] + quad[2][1]),
+                (quad[0][0] - quad[3][0]) * (quad[0][1] + quad[3][1])]
         return np.sum(edge) / 2.
-        
+
     def nms(self, dets):
         if self.is_python35:
             import lanms
@@ -141,7 +151,7 @@ class SASTPostProcess(object):
         """
         Cluster pixels in tcl_map based on quads.
         """
-        instance_count = quads.shape[0] + 1 # contain background
+        instance_count = quads.shape[0] + 1  # contain background
         instance_label_map = np.zeros(tcl_map.shape[:2], dtype=np.int32)
         if instance_count == 1:
             return instance_count, instance_label_map
@@ -149,18 +159,19 @@ class SASTPostProcess(object):
         # predict text center
         xy_text = np.argwhere(tcl_map[:, :, 0] > tcl_map_thresh)
         n = xy_text.shape[0]
-        xy_text = xy_text[:, ::-1] # (n, 2)
-        tco = tco_map[xy_text[:, 1], xy_text[:, 0], :] # (n, 2)
+        xy_text = xy_text[:, ::-1]  # (n, 2)
+        tco = tco_map[xy_text[:, 1], xy_text[:, 0], :]  # (n, 2)
         pred_tc = xy_text - tco
-        
+
         # get gt text center
         m = quads.shape[0]
-        gt_tc = np.mean(quads, axis=1) # (m, 2)
+        gt_tc = np.mean(quads, axis=1)  # (m, 2)
 
-        pred_tc_tile = np.tile(pred_tc[:, np.newaxis, :], (1, m, 1)) # (n, m, 2)
-        gt_tc_tile = np.tile(gt_tc[np.newaxis, :, :], (n, 1, 1)) # (n, m, 2)
-        dist_mat = np.linalg.norm(pred_tc_tile - gt_tc_tile, axis=2) # (n, m)
-        xy_text_assign = np.argmin(dist_mat, axis=1) + 1 # (n,)
+        pred_tc_tile = np.tile(pred_tc[:, np.newaxis, :],
+                               (1, m, 1))  # (n, m, 2)
+        gt_tc_tile = np.tile(gt_tc[np.newaxis, :, :], (n, 1, 1))  # (n, m, 2)
+        dist_mat = np.linalg.norm(pred_tc_tile - gt_tc_tile, axis=2)  # (n, m)
+        xy_text_assign = np.argmin(dist_mat, axis=1) + 1  # (n,)
 
         instance_label_map[xy_text[:, 1], xy_text[:, 0]] = xy_text_assign
         return instance_count, instance_label_map
@@ -169,26 +180,47 @@ class SASTPostProcess(object):
         """
         Estimate sample points number.
         """
-        eh = (np.linalg.norm(quad[0] - quad[3]) + np.linalg.norm(quad[1] - quad[2])) / 2.0
-        ew = (np.linalg.norm(quad[0] - quad[1]) + np.linalg.norm(quad[2] - quad[3])) / 2.0
+        eh = (np.linalg.norm(quad[0] - quad[3]) +
+              np.linalg.norm(quad[1] - quad[2])) / 2.0
+        ew = (np.linalg.norm(quad[0] - quad[1]) +
+              np.linalg.norm(quad[2] - quad[3])) / 2.0
 
         dense_sample_pts_num = max(2, int(ew))
-        dense_xy_center_line = xy_text[np.linspace(0, xy_text.shape[0] - 1, dense_sample_pts_num,
-                                                endpoint=True, dtype=np.float32).astype(np.int32)]
-
-        dense_xy_center_line_diff = dense_xy_center_line[1:] - dense_xy_center_line[:-1]
-        estimate_arc_len = np.sum(np.linalg.norm(dense_xy_center_line_diff, axis=1))
+        dense_xy_center_line = xy_text[np.linspace(
+            0,
+            xy_text.shape[0] - 1,
+            dense_sample_pts_num,
+            endpoint=True,
+            dtype=np.float32).astype(np.int32)]
+
+        dense_xy_center_line_diff = dense_xy_center_line[
+            1:] - dense_xy_center_line[:-1]
+        estimate_arc_len = np.sum(
+            np.linalg.norm(
+                dense_xy_center_line_diff, axis=1))
 
         sample_pts_num = max(2, int(estimate_arc_len / eh))
         return sample_pts_num
 
-    def detect_sast(self, tcl_map, tvo_map, tbo_map, tco_map, ratio_w, ratio_h, src_w, src_h, 
-                shrink_ratio_of_width=0.3, tcl_map_thresh=0.5, offset_expand=1.0, out_strid=4.0):
+    def detect_sast(self,
+                    tcl_map,
+                    tvo_map,
+                    tbo_map,
+                    tco_map,
+                    ratio_w,
+                    ratio_h,
+                    src_w,
+                    src_h,
+                    shrink_ratio_of_width=0.3,
+                    tcl_map_thresh=0.5,
+                    offset_expand=1.0,
+                    out_strid=4.0):
         """
         first resize the tcl_map, tvo_map and tbo_map to the input_size, then restore the polys
         """
         # restore quad
-        scores, quads, xy_text = self.restore_quad(tcl_map, tcl_map_thresh, tvo_map)
+        scores, quads, xy_text = self.restore_quad(tcl_map, tcl_map_thresh,
+                                                   tvo_map)
         dets = np.hstack((quads, scores)).astype(np.float32, copy=False)
         dets = self.nms(dets)
         if dets.shape[0] == 0:
@@ -202,7 +234,8 @@ class SASTPostProcess(object):
 
         # instance segmentation
         # instance_count, instance_label_map = cv2.connectedComponents(tcl_map.astype(np.uint8), connectivity=8)
-        instance_count, instance_label_map = self.cluster_by_quads_tco(tcl_map, tcl_map_thresh, quads, tco_map)
+        instance_count, instance_label_map = self.cluster_by_quads_tco(
+            tcl_map, tcl_map_thresh, quads, tco_map)
 
         # restore single poly with tcl instance.
         poly_list = []
@@ -212,10 +245,10 @@ class SASTPostProcess(object):
             q_area = quad_areas[instance_idx - 1]
             if q_area < 5:
                 continue
-            
+
             #
-            len1 = float(np.linalg.norm(quad[0] -quad[1]))
-            len2 = float(np.linalg.norm(quad[1] -quad[2]))
+            len1 = float(np.linalg.norm(quad[0] - quad[1]))
+            len2 = float(np.linalg.norm(quad[1] - quad[2]))
             min_len = min(len1, len2)
             if min_len < 3:
                 continue
@@ -225,16 +258,18 @@ class SASTPostProcess(object):
                 continue
 
             # filter low confidence instance
-            xy_text_scores = tcl_map[xy_text[:, 1], xy_text[:, 0], 0] 
+            xy_text_scores = tcl_map[xy_text[:, 1], xy_text[:, 0], 0]
             if np.sum(xy_text_scores) / quad_areas[instance_idx - 1] < 0.1:
-            # if np.sum(xy_text_scores) / quad_areas[instance_idx - 1] < 0.05:
+                # if np.sum(xy_text_scores) / quad_areas[instance_idx - 1] < 0.05:
                 continue
 
             # sort xy_text
-            left_center_pt = np.array([[(quad[0, 0] + quad[-1, 0]) / 2.0,
-                                        (quad[0, 1] + quad[-1, 1]) / 2.0]]) # (1, 2)
-            right_center_pt = np.array([[(quad[1, 0] + quad[2, 0]) / 2.0,
-                                        (quad[1, 1] + quad[2, 1]) / 2.0]]) # (1, 2)
+            left_center_pt = np.array(
+                [[(quad[0, 0] + quad[-1, 0]) / 2.0,
+                  (quad[0, 1] + quad[-1, 1]) / 2.0]])  # (1, 2)
+            right_center_pt = np.array(
+                [[(quad[1, 0] + quad[2, 0]) / 2.0,
+                  (quad[1, 1] + quad[2, 1]) / 2.0]])  # (1, 2)
             proj_unit_vec = (right_center_pt - left_center_pt) / \
                             (np.linalg.norm(right_center_pt - left_center_pt) + 1e-6)
             proj_value = np.sum(xy_text * proj_unit_vec, axis=1)
@@ -245,33 +280,45 @@ class SASTPostProcess(object):
                 sample_pts_num = self.estimate_sample_pts_num(quad, xy_text)
             else:
                 sample_pts_num = self.sample_pts_num
-            xy_center_line = xy_text[np.linspace(0, xy_text.shape[0] - 1, sample_pts_num,
-                                                endpoint=True, dtype=np.float32).astype(np.int32)]
+            xy_center_line = xy_text[np.linspace(
+                0,
+                xy_text.shape[0] - 1,
+                sample_pts_num,
+                endpoint=True,
+                dtype=np.float32).astype(np.int32)]
 
             point_pair_list = []
             for x, y in xy_center_line:
                 # get corresponding offset
                 offset = tbo_map[y, x, :].reshape(2, 2)
                 if offset_expand != 1.0:
-                    offset_length = np.linalg.norm(offset, axis=1, keepdims=True)
-                    expand_length = np.clip(offset_length * (offset_expand - 1), a_min=0.5, a_max=3.0)
+                    offset_length = np.linalg.norm(
+                        offset, axis=1, keepdims=True)
+                    expand_length = np.clip(
+                        offset_length * (offset_expand - 1),
+                        a_min=0.5,
+                        a_max=3.0)
                     offset_detal = offset / offset_length * expand_length
-                    offset = offset + offset_detal                
-                # original point
+                    offset = offset + offset_detal
+                    # original point
                 ori_yx = np.array([y, x], dtype=np.float32)
-                point_pair = (ori_yx +  offset)[:, ::-1]* out_strid / np.array([ratio_w, ratio_h]).reshape(-1, 2) 
+                point_pair = (ori_yx + offset)[:, ::-1] * out_strid / np.array(
+                    [ratio_w, ratio_h]).reshape(-1, 2)
                 point_pair_list.append(point_pair)
 
             # ndarry: (x, 2), expand poly along width
             detected_poly = self.point_pair2poly(point_pair_list)
-            detected_poly = self.expand_poly_along_width(detected_poly, shrink_ratio_of_width)
-            detected_poly[:, 0] = np.clip(detected_poly[:, 0], a_min=0, a_max=src_w)
-            detected_poly[:, 1] = np.clip(detected_poly[:, 1], a_min=0, a_max=src_h)
+            detected_poly = self.expand_poly_along_width(detected_poly,
+                                                         shrink_ratio_of_width)
+            detected_poly[:, 0] = np.clip(
+                detected_poly[:, 0], a_min=0, a_max=src_w)
+            detected_poly[:, 1] = np.clip(
+                detected_poly[:, 1], a_min=0, a_max=src_h)
             poly_list.append(detected_poly)
 
         return poly_list
 
-    def __call__(self, outs_dict, shape_list):                
+    def __call__(self, outs_dict, shape_list):
         score_list = outs_dict['f_score']
         border_list = outs_dict['f_border']
         tvo_list = outs_dict['f_tvo']
@@ -281,20 +328,28 @@ class SASTPostProcess(object):
             border_list = border_list.numpy()
             tvo_list = tvo_list.numpy()
             tco_list = tco_list.numpy()
-                    
+
         img_num = len(shape_list)
         poly_lists = []
         for ino in range(img_num):
-            p_score = score_list[ino].transpose((1,2,0))
-            p_border = border_list[ino].transpose((1,2,0))
-            p_tvo = tvo_list[ino].transpose((1,2,0))
-            p_tco = tco_list[ino].transpose((1,2,0))
+            p_score = score_list[ino].transpose((1, 2, 0))
+            p_border = border_list[ino].transpose((1, 2, 0))
+            p_tvo = tvo_list[ino].transpose((1, 2, 0))
+            p_tco = tco_list[ino].transpose((1, 2, 0))
             src_h, src_w, ratio_h, ratio_w = shape_list[ino]
 
-            poly_list = self.detect_sast(p_score, p_tvo, p_border, p_tco, ratio_w, ratio_h, src_w, src_h, 
-                                         shrink_ratio_of_width=self.shrink_ratio_of_width, 
-                                         tcl_map_thresh=self.tcl_map_thresh, offset_expand=self.expand_scale)
+            poly_list = self.detect_sast(
+                p_score,
+                p_tvo,
+                p_border,
+                p_tco,
+                ratio_w,
+                ratio_h,
+                src_w,
+                src_h,
+                shrink_ratio_of_width=self.shrink_ratio_of_width,
+                tcl_map_thresh=self.tcl_map_thresh,
+                offset_expand=self.expand_scale)
             poly_lists.append({'points': np.array(poly_list)})
 
         return poly_lists
-

ppocr/utils/e2e_metric/polygon_fast.py

+# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+import numpy as np
+from shapely.geometry import Polygon
+"""
+:param det_x: [1, N] Xs of detection's vertices 
+:param det_y: [1, N] Ys of detection's vertices
+:param gt_x: [1, N] Xs of groundtruth's vertices
+:param gt_y: [1, N] Ys of groundtruth's vertices
+
+##############
+All the calculation of 'AREA' in this script is handled by:
+1) First generating a binary mask with the polygon area filled up with 1's
+2) Summing up all the 1's
+"""
+
+
+def area(x, y):
+    polygon = Polygon(np.stack([x, y], axis=1))
+    return float(polygon.area)
+
+
+def approx_area_of_intersection(det_x, det_y, gt_x, gt_y):
+    """
+    This helper determine if both polygons are intersecting with each others with an approximation method.
+    Area of intersection represented by the minimum bounding rectangular [xmin, ymin, xmax, ymax]
+    """
+    det_ymax = np.max(det_y)
+    det_xmax = np.max(det_x)
+    det_ymin = np.min(det_y)
+    det_xmin = np.min(det_x)
+
+    gt_ymax = np.max(gt_y)
+    gt_xmax = np.max(gt_x)
+    gt_ymin = np.min(gt_y)
+    gt_xmin = np.min(gt_x)
+
+    all_min_ymax = np.minimum(det_ymax, gt_ymax)
+    all_max_ymin = np.maximum(det_ymin, gt_ymin)
+
+    intersect_heights = np.maximum(0.0, (all_min_ymax - all_max_ymin))
+
+    all_min_xmax = np.minimum(det_xmax, gt_xmax)
+    all_max_xmin = np.maximum(det_xmin, gt_xmin)
+    intersect_widths = np.maximum(0.0, (all_min_xmax - all_max_xmin))
+
+    return intersect_heights * intersect_widths
+
+
+def area_of_intersection(det_x, det_y, gt_x, gt_y):
+    p1 = Polygon(np.stack([det_x, det_y], axis=1)).buffer(0)
+    p2 = Polygon(np.stack([gt_x, gt_y], axis=1)).buffer(0)
+    return float(p1.intersection(p2).area)
+
+
+def area_of_union(det_x, det_y, gt_x, gt_y):
+    p1 = Polygon(np.stack([det_x, det_y], axis=1)).buffer(0)
+    p2 = Polygon(np.stack([gt_x, gt_y], axis=1)).buffer(0)
+    return float(p1.union(p2).area)
+
+
+def iou(det_x, det_y, gt_x, gt_y):
+    return area_of_intersection(det_x, det_y, gt_x, gt_y) / (
+        area_of_union(det_x, det_y, gt_x, gt_y) + 1.0)
+
+
+def iod(det_x, det_y, gt_x, gt_y):
+    """
+    This helper determine the fraction of intersection area over detection area
+    """
+    return area_of_intersection(det_x, det_y, gt_x, gt_y) / (
+        area(det_x, det_y) + 1.0)

ppocr/utils/e2e_utils/extract_batchsize.py

+import paddle
+import numpy as np
+import copy
+
+
+def org_tcl_rois(batch_size, pos_lists, pos_masks, label_lists, tcl_bs):
+    """
+    """
+    pos_lists_, pos_masks_, label_lists_ = [], [], []
+    img_bs = batch_size
+    ngpu = int(batch_size / img_bs)
+    img_ids = np.array(pos_lists, dtype=np.int32)[:, 0, 0].copy()
+    pos_lists_split, pos_masks_split, label_lists_split = [], [], []
+    for i in range(ngpu):
+        pos_lists_split.append([])
+        pos_masks_split.append([])
+        label_lists_split.append([])
+
+    for i in range(img_ids.shape[0]):
+        img_id = img_ids[i]
+        gpu_id = int(img_id / img_bs)
+        img_id = img_id % img_bs
+        pos_list = pos_lists[i].copy()
+        pos_list[:, 0] = img_id
+        pos_lists_split[gpu_id].append(pos_list)
+        pos_masks_split[gpu_id].append(pos_masks[i].copy())
+        label_lists_split[gpu_id].append(copy.deepcopy(label_lists[i]))
+    # repeat or delete
+    for i in range(ngpu):
+        vp_len = len(pos_lists_split[i])
+        if vp_len <= tcl_bs:
+            for j in range(0, tcl_bs - vp_len):
+                pos_list = pos_lists_split[i][j].copy()
+                pos_lists_split[i].append(pos_list)
+                pos_mask = pos_masks_split[i][j].copy()
+                pos_masks_split[i].append(pos_mask)
+                label_list = copy.deepcopy(label_lists_split[i][j])
+                label_lists_split[i].append(label_list)
+        else:
+            for j in range(0, vp_len - tcl_bs):
+                c_len = len(pos_lists_split[i])
+                pop_id = np.random.permutation(c_len)[0]
+                pos_lists_split[i].pop(pop_id)
+                pos_masks_split[i].pop(pop_id)
+                label_lists_split[i].pop(pop_id)
+    # merge
+    for i in range(ngpu):
+        pos_lists_.extend(pos_lists_split[i])
+        pos_masks_.extend(pos_masks_split[i])
+        label_lists_.extend(label_lists_split[i])
+    return pos_lists_, pos_masks_, label_lists_
+
+
+def pre_process(label_list, pos_list, pos_mask, max_text_length, max_text_nums,
+                pad_num, tcl_bs):
+    label_list = label_list.numpy()
+    batch, _, _, _ = label_list.shape
+    pos_list = pos_list.numpy()
+    pos_mask = pos_mask.numpy()
+    pos_list_t = []
+    pos_mask_t = []
+    label_list_t = []
+    for i in range(batch):
+        for j in range(max_text_nums):
+            if pos_mask[i, j].any():
+                pos_list_t.append(pos_list[i][j])
+                pos_mask_t.append(pos_mask[i][j])
+                label_list_t.append(label_list[i][j])
+    pos_list, pos_mask, label_list = org_tcl_rois(batch, pos_list_t, pos_mask_t,
+                                                  label_list_t, tcl_bs)
+    label = []
+    tt = [l.tolist() for l in label_list]
+    for i in range(tcl_bs):
+        k = 0
+        for j in range(max_text_length):
+            if tt[i][j][0] != pad_num:
+                k += 1
+            else:
+                break
+        label.append(k)
+    label = paddle.to_tensor(label)
+    label = paddle.cast(label, dtype='int64')
+    pos_list = paddle.to_tensor(pos_list)
+    pos_mask = paddle.to_tensor(pos_mask)
+    label_list = paddle.squeeze(paddle.to_tensor(label_list), axis=2)
+    label_list = paddle.cast(label_list, dtype='int32')
+    return pos_list, pos_mask, label_list, label

ppocr/utils/e2e_utils/extract_textpoint.py

+# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+"""Contains various CTC decoders."""
+from __future__ import absolute_import
+from __future__ import division
+from __future__ import print_function
+
+import cv2
+import math
+
+import numpy as np
+from itertools import groupby
+from skimage.morphology._skeletonize import thin
+
+
+def get_dict(character_dict_path):
+    character_str = ""
+    with open(character_dict_path, "rb") as fin:
+        lines = fin.readlines()
+        for line in lines:
+            line = line.decode('utf-8').strip("\n").strip("\r\n")
+            character_str += line
+        dict_character = list(character_str)
+    return dict_character
+
+
+def softmax(logits):
+    """
+    logits: N x d
+    """
+    max_value = np.max(logits, axis=1, keepdims=True)
+    exp = np.exp(logits - max_value)
+    exp_sum = np.sum(exp, axis=1, keepdims=True)
+    dist = exp / exp_sum
+    return dist
+
+
+def get_keep_pos_idxs(labels, remove_blank=None):
+    """
+    Remove duplicate and get pos idxs of keep items.
+    The value of keep_blank should be [None, 95].
+    """
+    duplicate_len_list = []
+    keep_pos_idx_list = []
+    keep_char_idx_list = []
+    for k, v_ in groupby(labels):
+        current_len = len(list(v_))
+        if k != remove_blank:
+            current_idx = int(sum(duplicate_len_list) + current_len // 2)
+            keep_pos_idx_list.append(current_idx)
+            keep_char_idx_list.append(k)
+        duplicate_len_list.append(current_len)
+    return keep_char_idx_list, keep_pos_idx_list
+
+
+def remove_blank(labels, blank=0):
+    new_labels = [x for x in labels if x != blank]
+    return new_labels
+
+
+def insert_blank(labels, blank=0):
+    new_labels = [blank]
+    for l in labels:
+        new_labels += [l, blank]
+    return new_labels
+
+
+def ctc_greedy_decoder(probs_seq, blank=95, keep_blank_in_idxs=True):
+    """
+    CTC greedy (best path) decoder.
+    """
+    raw_str = np.argmax(np.array(probs_seq), axis=1)
+    remove_blank_in_pos = None if keep_blank_in_idxs else blank
+    dedup_str, keep_idx_list = get_keep_pos_idxs(
+        raw_str, remove_blank=remove_blank_in_pos)
+    dst_str = remove_blank(dedup_str, blank=blank)
+    return dst_str, keep_idx_list
+
+
+def instance_ctc_greedy_decoder(gather_info,
+                                logits_map,
+                                keep_blank_in_idxs=True):
+    """
+    gather_info: [[x, y], [x, y] ...]
+    logits_map: H x W X (n_chars + 1)
+    """
+    _, _, C = logits_map.shape
+    ys, xs = zip(*gather_info)
+    logits_seq = logits_map[list(ys), list(xs)]  # n x 96
+    probs_seq = softmax(logits_seq)
+    dst_str, keep_idx_list = ctc_greedy_decoder(
+        probs_seq, blank=C - 1, keep_blank_in_idxs=keep_blank_in_idxs)
+    keep_gather_list = [gather_info[idx] for idx in keep_idx_list]
+    return dst_str, keep_gather_list
+
+
+def ctc_decoder_for_image(gather_info_list, logits_map,
+                          keep_blank_in_idxs=True):
+    """
+    CTC decoder using multiple processes.
+    """
+    decoder_results = []
+    for gather_info in gather_info_list:
+        res = instance_ctc_greedy_decoder(
+            gather_info, logits_map, keep_blank_in_idxs=keep_blank_in_idxs)
+        decoder_results.append(res)
+    return decoder_results
+
+
+def sort_with_direction(pos_list, f_direction):
+    """
+    f_direction: h x w x 2
+    pos_list: [[y, x], [y, x], [y, x] ...]
+    """
+
+    def sort_part_with_direction(pos_list, point_direction):
+        pos_list = np.array(pos_list).reshape(-1, 2)
+        point_direction = np.array(point_direction).reshape(-1, 2)
+        average_direction = np.mean(point_direction, axis=0, keepdims=True)
+        pos_proj_leng = np.sum(pos_list * average_direction, axis=1)
+        sorted_list = pos_list[np.argsort(pos_proj_leng)].tolist()
+        sorted_direction = point_direction[np.argsort(pos_proj_leng)].tolist()
+        return sorted_list, sorted_direction
+
+    pos_list = np.array(pos_list).reshape(-1, 2)
+    point_direction = f_direction[pos_list[:, 0], pos_list[:, 1]]  # x, y
+    point_direction = point_direction[:, ::-1]  # x, y -> y, x
+    sorted_point, sorted_direction = sort_part_with_direction(pos_list,
+                                                              point_direction)
+
+    point_num = len(sorted_point)
+    if point_num >= 16:
+        middle_num = point_num // 2
+        first_part_point = sorted_point[:middle_num]
+        first_point_direction = sorted_direction[:middle_num]
+        sorted_fist_part_point, sorted_fist_part_direction = sort_part_with_direction(
+            first_part_point, first_point_direction)
+
+        last_part_point = sorted_point[middle_num:]
+        last_point_direction = sorted_direction[middle_num:]
+        sorted_last_part_point, sorted_last_part_direction = sort_part_with_direction(
+            last_part_point, last_point_direction)
+        sorted_point = sorted_fist_part_point + sorted_last_part_point
+        sorted_direction = sorted_fist_part_direction + sorted_last_part_direction
+
+    return sorted_point, np.array(sorted_direction)
+
+
+def add_id(pos_list, image_id=0):
+    """
+    Add id for gather feature, for inference.
+    """
+    new_list = []
+    for item in pos_list:
+        new_list.append((image_id, item[0], item[1]))
+    return new_list
+
+
+def sort_and_expand_with_direction(pos_list, f_direction):
+    """
+    f_direction: h x w x 2
+    pos_list: [[y, x], [y, x], [y, x] ...]
+    """
+    h, w, _ = f_direction.shape
+    sorted_list, point_direction = sort_with_direction(pos_list, f_direction)
+
+    # expand along
+    point_num = len(sorted_list)
+    sub_direction_len = max(point_num // 3, 2)
+    left_direction = point_direction[:sub_direction_len, :]
+    right_dirction = point_direction[point_num - sub_direction_len:, :]
+
+    left_average_direction = -np.mean(left_direction, axis=0, keepdims=True)
+    left_average_len = np.linalg.norm(left_average_direction)
+    left_start = np.array(sorted_list[0])
+    left_step = left_average_direction / (left_average_len + 1e-6)
+
+    right_average_direction = np.mean(right_dirction, axis=0, keepdims=True)
+    right_average_len = np.linalg.norm(right_average_direction)
+    right_step = right_average_direction / (right_average_len + 1e-6)
+    right_start = np.array(sorted_list[-1])
+
+    append_num = max(
+        int((left_average_len + right_average_len) / 2.0 * 0.15), 1)
+    left_list = []
+    right_list = []
+    for i in range(append_num):
+        ly, lx = np.round(left_start + left_step * (i + 1)).flatten().astype(
+            'int32').tolist()
+        if ly < h and lx < w and (ly, lx) not in left_list:
+            left_list.append((ly, lx))
+        ry, rx = np.round(right_start + right_step * (i + 1)).flatten().astype(
+            'int32').tolist()
+        if ry < h and rx < w and (ry, rx) not in right_list:
+            right_list.append((ry, rx))
+
+    all_list = left_list[::-1] + sorted_list + right_list
+    return all_list
+
+
+def sort_and_expand_with_direction_v2(pos_list, f_direction, binary_tcl_map):
+    """
+    f_direction: h x w x 2
+    pos_list: [[y, x], [y, x], [y, x] ...]
+    binary_tcl_map: h x w
+    """
+    h, w, _ = f_direction.shape
+    sorted_list, point_direction = sort_with_direction(pos_list, f_direction)
+
+    # expand along
+    point_num = len(sorted_list)
+    sub_direction_len = max(point_num // 3, 2)
+    left_direction = point_direction[:sub_direction_len, :]
+    right_dirction = point_direction[point_num - sub_direction_len:, :]
+
+    left_average_direction = -np.mean(left_direction, axis=0, keepdims=True)
+    left_average_len = np.linalg.norm(left_average_direction)
+    left_start = np.array(sorted_list[0])
+    left_step = left_average_direction / (left_average_len + 1e-6)
+
+    right_average_direction = np.mean(right_dirction, axis=0, keepdims=True)
+    right_average_len = np.linalg.norm(right_average_direction)
+    right_step = right_average_direction / (right_average_len + 1e-6)
+    right_start = np.array(sorted_list[-1])
+
+    append_num = max(
+        int((left_average_len + right_average_len) / 2.0 * 0.15), 1)
+    max_append_num = 2 * append_num
+
+    left_list = []
+    right_list = []
+    for i in range(max_append_num):
+        ly, lx = np.round(left_start + left_step * (i + 1)).flatten().astype(
+            'int32').tolist()
+        if ly < h and lx < w and (ly, lx) not in left_list:
+            if binary_tcl_map[ly, lx] > 0.5:
+                left_list.append((ly, lx))
+            else:
+                break
+
+    for i in range(max_append_num):
+        ry, rx = np.round(right_start + right_step * (i + 1)).flatten().astype(
+            'int32').tolist()
+        if ry < h and rx < w and (ry, rx) not in right_list:
+            if binary_tcl_map[ry, rx] > 0.5:
+                right_list.append((ry, rx))
+            else:
+                break
+
+    all_list = left_list[::-1] + sorted_list + right_list
+    return all_list
+
+
+def generate_pivot_list_curved(p_score,
+                               p_char_maps,
+                               f_direction,
+                               score_thresh=0.5,
+                               is_expand=True,
+                               is_backbone=False,
+                               image_id=0):
+    """
+    return center point and end point of TCL instance; filter with the char maps;
+    """
+    p_score = p_score[0]
+    f_direction = f_direction.transpose(1, 2, 0)
+    p_tcl_map = (p_score > score_thresh) * 1.0
+    skeleton_map = thin(p_tcl_map)
+    instance_count, instance_label_map = cv2.connectedComponents(
+        skeleton_map.astype(np.uint8), connectivity=8)
+
+    # get TCL Instance
+    all_pos_yxs = []
+    center_pos_yxs = []
+    end_points_yxs = []
+    instance_center_pos_yxs = []
+    if instance_count > 0:
+        for instance_id in range(1, instance_count):
+            pos_list = []
+            ys, xs = np.where(instance_label_map == instance_id)
+            pos_list = list(zip(ys, xs))
+
+            ### FIX-ME, eliminate outlier
+            if len(pos_list) < 3:
+                continue
+
+            if is_expand:
+                pos_list_sorted = sort_and_expand_with_direction_v2(
+                    pos_list, f_direction, p_tcl_map)
+            else:
+                pos_list_sorted, _ = sort_with_direction(pos_list, f_direction)
+            all_pos_yxs.append(pos_list_sorted)
+
+    # use decoder to filter backgroud points.
+    p_char_maps = p_char_maps.transpose([1, 2, 0])
+    decode_res = ctc_decoder_for_image(
+        all_pos_yxs, logits_map=p_char_maps, keep_blank_in_idxs=True)
+    for decoded_str, keep_yxs_list in decode_res:
+        if is_backbone:
+            keep_yxs_list_with_id = add_id(keep_yxs_list, image_id=image_id)
+            instance_center_pos_yxs.append(keep_yxs_list_with_id)
+        else:
+            end_points_yxs.extend((keep_yxs_list[0], keep_yxs_list[-1]))
+            center_pos_yxs.extend(keep_yxs_list)
+
+    if is_backbone:
+        return instance_center_pos_yxs
+    else:
+        return center_pos_yxs, end_points_yxs
+
+
+def generate_pivot_list_horizontal(p_score,
+                                   p_char_maps,
+                                   f_direction,
+                                   score_thresh=0.5,
+                                   is_backbone=False,
+                                   image_id=0):
+    """
+    return center point and end point of TCL instance; filter with the char maps;
+    """
+    p_score = p_score[0]
+    f_direction = f_direction.transpose(1, 2, 0)
+    p_tcl_map_bi = (p_score > score_thresh) * 1.0
+    instance_count, instance_label_map = cv2.connectedComponents(
+        p_tcl_map_bi.astype(np.uint8), connectivity=8)
+
+    # get TCL Instance
+    all_pos_yxs = []
+    center_pos_yxs = []
+    end_points_yxs = []
+    instance_center_pos_yxs = []
+
+    if instance_count > 0:
+        for instance_id in range(1, instance_count):
+            pos_list = []
+            ys, xs = np.where(instance_label_map == instance_id)
+            pos_list = list(zip(ys, xs))
+
+            ### FIX-ME, eliminate outlier
+            if len(pos_list) < 5:
+                continue
+
+            # add rule here
+            main_direction = extract_main_direction(pos_list,
+                                                    f_direction)  # y x
+            reference_directin = np.array([0, 1]).reshape([-1, 2])  # y x
+            is_h_angle = abs(np.sum(
+                main_direction * reference_directin)) < math.cos(math.pi / 180 *
+                                                                 70)
+
+            point_yxs = np.array(pos_list)
+            max_y, max_x = np.max(point_yxs, axis=0)
+            min_y, min_x = np.min(point_yxs, axis=0)
+            is_h_len = (max_y - min_y) < 1.5 * (max_x - min_x)
+
+            pos_list_final = []
+            if is_h_len:
+                xs = np.unique(xs)
+                for x in xs:
+                    ys = instance_label_map[:, x].copy().reshape((-1, ))
+                    y = int(np.where(ys == instance_id)[0].mean())
+                    pos_list_final.append((y, x))
+            else:
+                ys = np.unique(ys)
+                for y in ys:
+                    xs = instance_label_map[y, :].copy().reshape((-1, ))
+                    x = int(np.where(xs == instance_id)[0].mean())
+                    pos_list_final.append((y, x))
+
+            pos_list_sorted, _ = sort_with_direction(pos_list_final,
+                                                     f_direction)
+            all_pos_yxs.append(pos_list_sorted)
+
+    # use decoder to filter backgroud points.
+    p_char_maps = p_char_maps.transpose([1, 2, 0])
+    decode_res = ctc_decoder_for_image(
+        all_pos_yxs, logits_map=p_char_maps, keep_blank_in_idxs=True)
+    for decoded_str, keep_yxs_list in decode_res:
+        if is_backbone:
+            keep_yxs_list_with_id = add_id(keep_yxs_list, image_id=image_id)
+            instance_center_pos_yxs.append(keep_yxs_list_with_id)
+        else:
+            end_points_yxs.extend((keep_yxs_list[0], keep_yxs_list[-1]))
+            center_pos_yxs.extend(keep_yxs_list)
+
+    if is_backbone:
+        return instance_center_pos_yxs
+    else:
+        return center_pos_yxs, end_points_yxs
+
+
+def generate_pivot_list(p_score,
+                        p_char_maps,
+                        f_direction,
+                        score_thresh=0.5,
+                        is_backbone=False,
+                        is_curved=True,
+                        image_id=0):
+    """
+    Warp all the function together.
+    """
+    if is_curved:
+        return generate_pivot_list_curved(
+            p_score,
+            p_char_maps,
+            f_direction,
+            score_thresh=score_thresh,
+            is_expand=True,
+            is_backbone=is_backbone,
+            image_id=image_id)
+    else:
+        return generate_pivot_list_horizontal(
+            p_score,
+            p_char_maps,
+            f_direction,
+            score_thresh=score_thresh,
+            is_backbone=is_backbone,
+            image_id=image_id)
+
+
+# for refine module
+def extract_main_direction(pos_list, f_direction):
+    """
+    f_direction: h x w x 2
+    pos_list: [[y, x], [y, x], [y, x] ...]
+    """
+    pos_list = np.array(pos_list)
+    point_direction = f_direction[pos_list[:, 0], pos_list[:, 1]]
+    point_direction = point_direction[:, ::-1]  # x, y -> y, x
+    average_direction = np.mean(point_direction, axis=0, keepdims=True)
+    average_direction = average_direction / (
+        np.linalg.norm(average_direction) + 1e-6)
+    return average_direction
+
+
+def sort_by_direction_with_image_id_deprecated(pos_list, f_direction):
+    """
+    f_direction: h x w x 2
+    pos_list: [[id, y, x], [id, y, x], [id, y, x] ...]
+    """
+    pos_list_full = np.array(pos_list).reshape(-1, 3)
+    pos_list = pos_list_full[:, 1:]
+    point_direction = f_direction[pos_list[:, 0], pos_list[:, 1]]  # x, y
+    point_direction = point_direction[:, ::-1]  # x, y -> y, x
+    average_direction = np.mean(point_direction, axis=0, keepdims=True)
+    pos_proj_leng = np.sum(pos_list * average_direction, axis=1)
+    sorted_list = pos_list_full[np.argsort(pos_proj_leng)].tolist()
+    return sorted_list
+
+
+def sort_by_direction_with_image_id(pos_list, f_direction):
+    """
+    f_direction: h x w x 2
+    pos_list: [[y, x], [y, x], [y, x] ...]
+    """
+
+    def sort_part_with_direction(pos_list_full, point_direction):
+        pos_list_full = np.array(pos_list_full).reshape(-1, 3)
+        pos_list = pos_list_full[:, 1:]
+        point_direction = np.array(point_direction).reshape(-1, 2)
+        average_direction = np.mean(point_direction, axis=0, keepdims=True)
+        pos_proj_leng = np.sum(pos_list * average_direction, axis=1)
+        sorted_list = pos_list_full[np.argsort(pos_proj_leng)].tolist()
+        sorted_direction = point_direction[np.argsort(pos_proj_leng)].tolist()
+        return sorted_list, sorted_direction
+
+    pos_list = np.array(pos_list).reshape(-1, 3)
+    point_direction = f_direction[pos_list[:, 1], pos_list[:, 2]]  # x, y
+    point_direction = point_direction[:, ::-1]  # x, y -> y, x
+    sorted_point, sorted_direction = sort_part_with_direction(pos_list,
+                                                              point_direction)
+
+    point_num = len(sorted_point)
+    if point_num >= 16:
+        middle_num = point_num // 2
+        first_part_point = sorted_point[:middle_num]
+        first_point_direction = sorted_direction[:middle_num]
+        sorted_fist_part_point, sorted_fist_part_direction = sort_part_with_direction(
+            first_part_point, first_point_direction)
+
+        last_part_point = sorted_point[middle_num:]
+        last_point_direction = sorted_direction[middle_num:]
+        sorted_last_part_point, sorted_last_part_direction = sort_part_with_direction(
+            last_part_point, last_point_direction)
+        sorted_point = sorted_fist_part_point + sorted_last_part_point
+        sorted_direction = sorted_fist_part_direction + sorted_last_part_direction
+
+    return sorted_point
+
+
+def generate_pivot_list_tt_inference(p_score,
+                                     p_char_maps,
+                                     f_direction,
+                                     score_thresh=0.5,
+                                     is_backbone=False,
+                                     is_curved=True,
+                                     image_id=0):
+    """
+    return center point and end point of TCL instance; filter with the char maps;
+    """
+    p_score = p_score[0]
+    f_direction = f_direction.transpose(1, 2, 0)
+    p_tcl_map = (p_score > score_thresh) * 1.0
+    skeleton_map = thin(p_tcl_map)
+    instance_count, instance_label_map = cv2.connectedComponents(
+        skeleton_map.astype(np.uint8), connectivity=8)
+
+    # get TCL Instance
+    all_pos_yxs = []
+    if instance_count > 0:
+        for instance_id in range(1, instance_count):
+            pos_list = []
+            ys, xs = np.where(instance_label_map == instance_id)
+            pos_list = list(zip(ys, xs))
+            ### FIX-ME, eliminate outlier
+            if len(pos_list) < 3:
+                continue
+            pos_list_sorted = sort_and_expand_with_direction_v2(
+                pos_list, f_direction, p_tcl_map)
+            pos_list_sorted_with_id = add_id(pos_list_sorted, image_id=image_id)
+            all_pos_yxs.append(pos_list_sorted_with_id)
+    return all_pos_yxs

ppocr/utils/e2e_utils/visual.py

+# Copyright (c) 2021 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+import numpy as np
+import cv2
+import time
+
+
+def resize_image(im, max_side_len=512):
+    """
+    resize image to a size multiple of max_stride which is required by the network
+    :param im: the resized image
+    :param max_side_len: limit of max image size to avoid out of memory in gpu
+    :return: the resized image and the resize ratio
+    """
+    h, w, _ = im.shape
+
+    resize_w = w
+    resize_h = h
+
+    if resize_h > resize_w:
+        ratio = float(max_side_len) / resize_h
+    else:
+        ratio = float(max_side_len) / resize_w
+
+    resize_h = int(resize_h * ratio)
+    resize_w = int(resize_w * ratio)
+
+    max_stride = 128
+    resize_h = (resize_h + max_stride - 1) // max_stride * max_stride
+    resize_w = (resize_w + max_stride - 1) // max_stride * max_stride
+    im = cv2.resize(im, (int(resize_w), int(resize_h)))
+    ratio_h = resize_h / float(h)
+    ratio_w = resize_w / float(w)
+
+    return im, (ratio_h, ratio_w)
+
+
+def resize_image_min(im, max_side_len=512):
+    """
+    """
+    h, w, _ = im.shape
+
+    resize_w = w
+    resize_h = h
+
+    if resize_h < resize_w:
+        ratio = float(max_side_len) / resize_h
+    else:
+        ratio = float(max_side_len) / resize_w
+
+    resize_h = int(resize_h * ratio)
+    resize_w = int(resize_w * ratio)
+
+    max_stride = 128
+    resize_h = (resize_h + max_stride - 1) // max_stride * max_stride
+    resize_w = (resize_w + max_stride - 1) // max_stride * max_stride
+    im = cv2.resize(im, (int(resize_w), int(resize_h)))
+    ratio_h = resize_h / float(h)
+    ratio_w = resize_w / float(w)
+    return im, (ratio_h, ratio_w)
+
+
+def resize_image_for_totaltext(im, max_side_len=512):
+    """
+    """
+    h, w, _ = im.shape
+
+    resize_w = w
+    resize_h = h
+    ratio = 1.25
+    if h * ratio > max_side_len:
+        ratio = float(max_side_len) / resize_h
+
+    resize_h = int(resize_h * ratio)
+    resize_w = int(resize_w * ratio)
+
+    max_stride = 128
+    resize_h = (resize_h + max_stride - 1) // max_stride * max_stride
+    resize_w = (resize_w + max_stride - 1) // max_stride * max_stride
+    im = cv2.resize(im, (int(resize_w), int(resize_h)))
+    ratio_h = resize_h / float(h)
+    ratio_w = resize_w / float(w)
+    return im, (ratio_h, ratio_w)
+
+
+def point_pair2poly(point_pair_list):
+    """
+    Transfer vertical point_pairs into poly point in clockwise.
+    """
+    pair_length_list = []
+    for point_pair in point_pair_list:
+        pair_length = np.linalg.norm(point_pair[0] - point_pair[1])
+        pair_length_list.append(pair_length)
+    pair_length_list = np.array(pair_length_list)
+    pair_info = (pair_length_list.max(), pair_length_list.min(),
+                 pair_length_list.mean())
+
+    point_num = len(point_pair_list) * 2
+    point_list = [0] * point_num
+    for idx, point_pair in enumerate(point_pair_list):
+        point_list[idx] = point_pair[0]
+        point_list[point_num - 1 - idx] = point_pair[1]
+    return np.array(point_list).reshape(-1, 2), pair_info
+
+
+def shrink_quad_along_width(quad, begin_width_ratio=0., end_width_ratio=1.):
+    """
+    Generate shrink_quad_along_width.
+    """
+    ratio_pair = np.array(
+        [[begin_width_ratio], [end_width_ratio]], dtype=np.float32)
+    p0_1 = quad[0] + (quad[1] - quad[0]) * ratio_pair
+    p3_2 = quad[3] + (quad[2] - quad[3]) * ratio_pair
+    return np.array([p0_1[0], p0_1[1], p3_2[1], p3_2[0]])
+
+
+def expand_poly_along_width(poly, shrink_ratio_of_width=0.3):
+    """
+    expand poly along width.
+    """
+    point_num = poly.shape[0]
+    left_quad = np.array(
+        [poly[0], poly[1], poly[-2], poly[-1]], dtype=np.float32)
+    left_ratio = -shrink_ratio_of_width * np.linalg.norm(left_quad[0] - left_quad[3]) / \
+                 (np.linalg.norm(left_quad[0] - left_quad[1]) + 1e-6)
+    left_quad_expand = shrink_quad_along_width(left_quad, left_ratio, 1.0)
+    right_quad = np.array(
+        [
+            poly[point_num // 2 - 2], poly[point_num // 2 - 1],
+            poly[point_num // 2], poly[point_num // 2 + 1]
+        ],
+        dtype=np.float32)
+    right_ratio = 1.0 + \
+                  shrink_ratio_of_width * np.linalg.norm(right_quad[0] - right_quad[3]) / \
+                  (np.linalg.norm(right_quad[0] - right_quad[1]) + 1e-6)
+    right_quad_expand = shrink_quad_along_width(right_quad, 0.0, right_ratio)
+    poly[0] = left_quad_expand[0]
+    poly[-1] = left_quad_expand[-1]
+    poly[point_num // 2 - 1] = right_quad_expand[1]
+    poly[point_num // 2] = right_quad_expand[2]
+    return poly
+
+
+def norm2(x, axis=None):
+    if axis:
+        return np.sqrt(np.sum(x**2, axis=axis))
+    return np.sqrt(np.sum(x**2))
+
+
+def cos(p1, p2):
+    return (p1 * p2).sum() / (norm2(p1) * norm2(p2))

ppocr/utils/utility.py

@@ -61,6 +61,7 @@ def get_image_file_list(img_file):
                 imgs_lists.append(file_path)
     if len(imgs_lists) == 0:
         raise Exception("not found any img file in {}".format(img_file))
+    imgs_lists = sorted(imgs_lists)
     return imgs_lists
 
 

requirements.txt

@@ -7,4 +7,5 @@ opencv-python==4.2.0.32
 tqdm
 numpy
 visualdl
-python-Levenshtein
+python-Levenshtein
\ No newline at end of file
+opencv-contrib-python
\ No newline at end of file

tools/infer/predict_e2e.py

+# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+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__, '../..')))
+
+os.environ["FLAGS_allocator_strategy"] = 'auto_growth'
+
+import cv2
+import numpy as np
+import time
+import sys
+
+import tools.infer.utility as utility
+from ppocr.utils.logging import get_logger
+from ppocr.utils.utility import get_image_file_list, check_and_read_gif
+from ppocr.data import create_operators, transform
+from ppocr.postprocess import build_post_process
+
+logger = get_logger()
+
+
+class TextE2E(object):
+    def __init__(self, args):
+        self.args = args
+        self.e2e_algorithm = args.e2e_algorithm
+        pre_process_list = [{
+            'E2EResizeForTest': {}
+        }, {
+            'NormalizeImage': {
+                'std': [0.229, 0.224, 0.225],
+                'mean': [0.485, 0.456, 0.406],
+                'scale': '1./255.',
+                'order': 'hwc'
+            }
+        }, {
+            'ToCHWImage': None
+        }, {
+            'KeepKeys': {
+                'keep_keys': ['image', 'shape']
+            }
+        }]
+        postprocess_params = {}
+        if self.e2e_algorithm == "PGNet":
+            pre_process_list[0] = {
+                'E2EResizeForTest': {
+                    'max_side_len': args.e2e_limit_side_len,
+                    'valid_set': 'totaltext'
+                }
+            }
+            postprocess_params['name'] = 'PGPostProcess'
+            postprocess_params["score_thresh"] = args.e2e_pgnet_score_thresh
+            postprocess_params["character_dict_path"] = args.e2e_char_dict_path
+            postprocess_params["valid_set"] = args.e2e_pgnet_valid_set
+            self.e2e_pgnet_polygon = args.e2e_pgnet_polygon
+        else:
+            logger.info("unknown e2e_algorithm:{}".format(self.e2e_algorithm))
+            sys.exit(0)
+
+        self.preprocess_op = create_operators(pre_process_list)
+        self.postprocess_op = build_post_process(postprocess_params)
+        self.predictor, self.input_tensor, self.output_tensors = utility.create_predictor(
+            args, 'e2e', logger)  # paddle.jit.load(args.det_model_dir)
+        # self.predictor.eval()
+
+    def clip_det_res(self, points, img_height, img_width):
+        for pno in range(points.shape[0]):
+            points[pno, 0] = int(min(max(points[pno, 0], 0), img_width - 1))
+            points[pno, 1] = int(min(max(points[pno, 1], 0), img_height - 1))
+        return points
+
+    def filter_tag_det_res_only_clip(self, dt_boxes, image_shape):
+        img_height, img_width = image_shape[0:2]
+        dt_boxes_new = []
+        for box in dt_boxes:
+            box = self.clip_det_res(box, img_height, img_width)
+            dt_boxes_new.append(box)
+        dt_boxes = np.array(dt_boxes_new)
+        return dt_boxes
+
+    def __call__(self, img):
+
+        ori_im = img.copy()
+        data = {'image': img}
+        data = transform(data, self.preprocess_op)
+        img, shape_list = data
+        if img is None:
+            return None, 0
+        img = np.expand_dims(img, axis=0)
+        shape_list = np.expand_dims(shape_list, axis=0)
+        img = img.copy()
+        starttime = time.time()
+
+        self.input_tensor.copy_from_cpu(img)
+        self.predictor.run()
+        outputs = []
+        for output_tensor in self.output_tensors:
+            output = output_tensor.copy_to_cpu()
+            outputs.append(output)
+
+        preds = {}
+        if self.e2e_algorithm == 'PGNet':
+            preds['f_border'] = outputs[0]
+            preds['f_char'] = outputs[1]
+            preds['f_direction'] = outputs[2]
+            preds['f_score'] = outputs[3]
+        else:
+            raise NotImplementedError
+        post_result = self.postprocess_op(preds, shape_list)
+        points, strs = post_result['points'], post_result['strs']
+        dt_boxes = self.filter_tag_det_res_only_clip(points, ori_im.shape)
+        elapse = time.time() - starttime
+        return dt_boxes, strs, elapse
+
+
+if __name__ == "__main__":
+    args = utility.parse_args()
+    image_file_list = get_image_file_list(args.image_dir)
+    text_detector = TextE2E(args)
+    count = 0
+    total_time = 0
+    draw_img_save = "./inference_results"
+    if not os.path.exists(draw_img_save):
+        os.makedirs(draw_img_save)
+    for image_file in image_file_list:
+        img, flag = check_and_read_gif(image_file)
+        if not flag:
+            img = cv2.imread(image_file)
+        if img is None:
+            logger.info("error in loading image:{}".format(image_file))
+            continue
+        points, strs, elapse = text_detector(img)
+        if count > 0:
+            total_time += elapse
+        count += 1
+        logger.info("Predict time of {}: {}".format(image_file, elapse))
+        src_im = utility.draw_e2e_res(points, strs, image_file)
+        img_name_pure = os.path.split(image_file)[-1]
+        img_path = os.path.join(draw_img_save,
+                                "e2e_res_{}".format(img_name_pure))
+        cv2.imwrite(img_path, src_im)
+        logger.info("The visualized image saved in {}".format(img_path))
+    if count > 1:
+        logger.info("Avg Time: {}".format(total_time / (count - 1)))

tools/infer/predict_rec.py

@@ -41,6 +41,7 @@ class TextRecognizer(object):
         self.character_type = args.rec_char_type
         self.rec_batch_num = args.rec_batch_num
         self.rec_algorithm = args.rec_algorithm
+        self.max_text_length = args.max_text_length
         postprocess_params = {
             'name': 'CTCLabelDecode',
             "character_type": args.rec_char_type,
@@ -186,8 +187,9 @@ class TextRecognizer(object):
                     norm_img = norm_img[np.newaxis, :]
                     norm_img_batch.append(norm_img)
                 else:
-                    norm_img = self.process_image_srn(
-                        img_list[indices[ino]], self.rec_image_shape, 8, 25)
+                    norm_img = self.process_image_srn(img_list[indices[ino]],
+                                                      self.rec_image_shape, 8,
+                                                      self.max_text_length)
                     encoder_word_pos_list = []
                     gsrm_word_pos_list = []
                     gsrm_slf_attn_bias1_list = []

tools/infer/predict_system.py

@@ -13,6 +13,7 @@
 # limitations under the License.
 import os
 import sys
+import subprocess
 
 __dir__ = os.path.dirname(os.path.abspath(__file__))
 sys.path.append(__dir__)
@@ -141,6 +142,7 @@ def sorted_boxes(dt_boxes):
 
 def main(args):
     image_file_list = get_image_file_list(args.image_dir)
+    image_file_list = image_file_list[args.process_id::args.total_process_num]
     text_sys = TextSystem(args)
     is_visualize = True
     font_path = args.vis_font_path
@@ -184,4 +186,18 @@ def main(args):
 
 
 if __name__ == "__main__":
-    main(utility.parse_args())
+    args = utility.parse_args()
+    if args.use_mp:
+        p_list = []
+        total_process_num = args.total_process_num
+        for process_id in range(total_process_num):
+            cmd = [sys.executable, "-u"] + sys.argv + [
+                "--process_id={}".format(process_id),
+                "--use_mp={}".format(False)
+            ]
+            p = subprocess.Popen(cmd, stdout=sys.stdout, stderr=sys.stdout)
+            p_list.append(p)
+        for p in p_list:
+            p.wait()
+    else:
+        main(args)

tools/infer/utility.py

@@ -74,6 +74,19 @@ def parse_args():
         "--vis_font_path", type=str, default="./doc/fonts/simfang.ttf")
     parser.add_argument("--drop_score", type=float, default=0.5)
 
+    # params for e2e
+    parser.add_argument("--e2e_algorithm", type=str, default='PGNet')
+    parser.add_argument("--e2e_model_dir", type=str)
+    parser.add_argument("--e2e_limit_side_len", type=float, default=768)
+    parser.add_argument("--e2e_limit_type", type=str, default='max')
+
+    # PGNet parmas
+    parser.add_argument("--e2e_pgnet_score_thresh", type=float, default=0.5)
+    parser.add_argument(
+        "--e2e_char_dict_path", type=str, default="./ppocr/utils/ic15_dict.txt")
+    parser.add_argument("--e2e_pgnet_valid_set", type=str, default='totaltext')
+    parser.add_argument("--e2e_pgnet_polygon", type=bool, default=True)
+
     # params for text classifier
     parser.add_argument("--use_angle_cls", type=str2bool, default=False)
     parser.add_argument("--cls_model_dir", type=str)
@@ -85,6 +98,10 @@ def parse_args():
     parser.add_argument("--enable_mkldnn", type=str2bool, default=False)
     parser.add_argument("--use_pdserving", type=str2bool, default=False)
 
+    parser.add_argument("--use_mp", type=str2bool, default=False)
+    parser.add_argument("--total_process_num", type=int, default=1)
+    parser.add_argument("--process_id", type=int, default=0)
+
     return parser.parse_args()
 
 
@@ -93,8 +110,10 @@ def create_predictor(args, mode, logger):
         model_dir = args.det_model_dir
     elif mode == 'cls':
         model_dir = args.cls_model_dir
-    else:
+    elif mode == 'rec':
         model_dir = args.rec_model_dir
+    else:
+        model_dir = args.e2e_model_dir
 
     if model_dir is None:
         logger.info("not find {} model file path {}".format(mode, model_dir))
@@ -148,6 +167,22 @@ def create_predictor(args, mode, logger):
     return predictor, input_tensor, output_tensors
 
 
+def draw_e2e_res(dt_boxes, strs, img_path):
+    src_im = cv2.imread(img_path)
+    for box, str in zip(dt_boxes, strs):
+        box = box.astype(np.int32).reshape((-1, 1, 2))
+        cv2.polylines(src_im, [box], True, color=(255, 255, 0), thickness=2)
+        cv2.putText(
+            src_im,
+            str,
+            org=(int(box[0, 0, 0]), int(box[0, 0, 1])),
+            fontFace=cv2.FONT_HERSHEY_COMPLEX,
+            fontScale=0.7,
+            color=(0, 255, 0),
+            thickness=1)
+    return src_im
+
+
 def draw_text_det_res(dt_boxes, img_path):
     src_im = cv2.imread(img_path)
     for box in dt_boxes:

tools/program.py

@@ -375,7 +375,8 @@ def preprocess(is_train=False):
 
     alg = config['Architecture']['algorithm']
     assert alg in [
-        'EAST', 'DB', 'SAST', 'Rosetta', 'CRNN', 'STARNet', 'RARE', 'SRN', 'CLS'
+        'EAST', 'DB', 'SAST', 'Rosetta', 'CRNN', 'STARNet', 'RARE', 'SRN',
+        'CLS', 'PGNet'
     ]
 
     device = 'gpu:{}'.format(dist.ParallelEnv().dev_id) if use_gpu else 'cpu'