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

doc/doc_ch/recognition.md

@@ -36,6 +36,7 @@ ln -sf <path/to/dataset> <path/to/paddle_ocr>/train_data/dataset
 * 数据下载
 
 若您本地没有数据集，可以在官网下载 [icdar2015](http://rrc.cvc.uab.es/?ch=4&com=downloads) 数据，用于快速验证。也可以参考[DTRB](https://github.com/clovaai/deep-text-recognition-benchmark#download-lmdb-dataset-for-traininig-and-evaluation-from-here)，下载 benchmark 所需的lmdb格式数据集。
+如果希望复现SRN的论文指标，需要下载离线[增广数据](https://pan.baidu.com/s/1-HSZ-ZVdqBF2HaBZ5pRAKA),提取码: y3ry。增广数据是由MJSynth和SynthText做旋转和扰动得到的。数据下载完成后请解压到 {your_path}/PaddleOCR/train_data/data_lmdb_release/training/ 路径下。
 
 <a name="自定义数据集"></a>
 * 使用自己数据集
@@ -132,7 +133,7 @@ word_dict.txt 每行有一个单字，将字符与数字索引映射在一起，
 您可以按需使用。
 
 目前的多语言模型仍处在demo阶段，会持续优化模型并补充语种，**非常欢迎您为我们提供其他语言的字典和字体**，
-如您愿意可将字典文件提交至 [dict](../../ppocr/utils/dict) 将语料文件提交至[corpus](../../ppocr/utils/corpus)，我们会在Repo中感谢您。
+如您愿意可将字典文件提交至 [dict](../../ppocr/utils/dict)，我们会在Repo中感谢您。
 
 - 自定义字典
 
@@ -200,6 +201,9 @@ PaddleOCR支持训练和评估交替进行, 可以在 `configs/rec/rec_icdar15_t
 | rec_mv3_none_none_ctc.yml |  Rosetta |   Mobilenet_v3 large 0.5 |  None   |  None |  ctc  |
 | rec_r34_vd_none_bilstm_ctc.yml |  CRNN |   Resnet34_vd |  None   |  BiLSTM |  ctc  |
 | rec_r34_vd_none_none_ctc.yml |  Rosetta |   Resnet34_vd |  None   |  None |  ctc  |
+| rec_mv3_tps_bilstm_att.yml |  CRNN |   Mobilenet_v3 |  TPS   |  BiLSTM |  att  |
+| rec_r34_vd_tps_bilstm_att.yml |  CRNN |   Resnet34_vd |  TPS   |  BiLSTM |  att  |
+| rec_r50fpn_vd_none_srn.yml    | SRN | Resnet50_fpn_vd    | None    | rnn | srn |
 
 训练中文数据，推荐使用[rec_chinese_lite_train_v2.0.yml](../../configs/rec/ch_ppocr_v2.0/rec_chinese_lite_train_v2.0.yml)，如您希望尝试其他算法在中文数据集上的效果，请参考下列说明修改配置文件：
 

doc/doc_en/algorithm_overview_en.md

@@ -42,8 +42,8 @@ PaddleOCR open-source text recognition algorithms list:
 - [x]  CRNN([paper](https://arxiv.org/abs/1507.05717))[7]
 - [x]  Rosetta([paper](https://arxiv.org/abs/1910.05085))[10]
 - [x]  STAR-Net([paper](http://www.bmva.org/bmvc/2016/papers/paper043/index.html))[11]
-- [ ]  RARE([paper](https://arxiv.org/abs/1603.03915v1))[12] coming soon
-- [ ]  SRN([paper](https://arxiv.org/abs/2003.12294))[5] coming soon
+- [x]  RARE([paper](https://arxiv.org/abs/1603.03915v1))[12]
+- [x]  SRN([paper](https://arxiv.org/abs/2003.12294))[5]
 
 Refer to [DTRB](https://arxiv.org/abs/1904.01906), the training and evaluation result of these above text recognition (using MJSynth and SynthText for training, evaluate on IIIT, SVT, IC03, IC13, IC15, SVTP, CUTE) is as follow:
 
@@ -55,5 +55,8 @@ Refer to [DTRB](https://arxiv.org/abs/1904.01906), the training and evaluation r
 |CRNN|MobileNetV3|79.97%|rec_mv3_none_bilstm_ctc|[Download link](https://paddleocr.bj.bcebos.com/dygraph_v2.0/en/rec_mv3_none_bilstm_ctc_v2.0_train.tar)|
 |StarNet|Resnet34_vd|84.44%|rec_r34_vd_tps_bilstm_ctc|[Download link](https://paddleocr.bj.bcebos.com/dygraph_v2.0/en/rec_r34_vd_tps_bilstm_ctc_v2.0_train.tar)|
 |StarNet|MobileNetV3|81.42%|rec_mv3_tps_bilstm_ctc|[Download link](https://paddleocr.bj.bcebos.com/dygraph_v2.0/en/rec_mv3_tps_bilstm_ctc_v2.0_train.tar)|
+|RARE|MobileNetV3|82.5%|rec_mv3_tps_bilstm_att |[Download link](https://paddleocr.bj.bcebos.com/dygraph_v2.0/en/rec_mv3_tps_bilstm_att_v2.0_train.tar)|
+|RARE|Resnet34_vd|83.6%|rec_r34_vd_tps_bilstm_att |[Download link](https://paddleocr.bj.bcebos.com/dygraph_v2.0/en/rec_r34_vd_tps_bilstm_att_v2.0_train.tar)|
+|SRN|Resnet50_vd_fpn| 88.52% | rec_r50fpn_vd_none_srn |[Download link](https://paddleocr.bj.bcebos.com/dygraph_v2.0/en/rec_r50_vd_srn_train.tar)|
 
 Please refer to the document for training guide and use of PaddleOCR text recognition algorithms [Text recognition model training/evaluation/prediction](./recognition_en.md)

doc/doc_en/angle_class_en.md

@@ -66,7 +66,7 @@ Start training:
 ```
 # Set PYTHONPATH path
 export PYTHONPATH=$PYTHONPATH:.
-# GPU training Support single card and multi-card training, specify the card number through --gpus. If your paddle version is less than 2.0rc1, please use '--selected_gpus'
+# GPU training Support single card and multi-card training, specify the card number through --gpus.
 # Start training, the following command has been written into the train.sh file, just modify the configuration file path in the file
 python3 -m paddle.distributed.launch --gpus '0,1,2,3,4,5,6,7'  tools/train.py -c configs/cls/cls_mv3.yml
 ```

doc/doc_en/detection_en.md

@@ -76,7 +76,7 @@ You can also use `-o` to change the training parameters without modifying the ym
 python3 tools/train.py -c configs/det/det_mv3_db.yml -o Optimizer.base_lr=0.0001
 
 # multi-GPU training
-# Set the GPU ID used by the '--gpus' parameter; If your paddle version is less than 2.0rc1, please use '--selected_gpus'
+# Set the GPU ID used by the '--gpus' parameter.
 python3 -m paddle.distributed.launch --gpus '0,1,2,3'  tools/train.py -c configs/det/det_mv3_db.yml -o Optimizer.base_lr=0.0001
 
 

doc/doc_en/inference_en.md

@@ -25,6 +25,7 @@ Next, we first introduce how to convert a trained model into an inference model,
 - [TEXT RECOGNITION MODEL INFERENCE](#RECOGNITION_MODEL_INFERENCE)
     - [1. LIGHTWEIGHT CHINESE MODEL](#LIGHTWEIGHT_RECOGNITION)
     - [2. CTC-BASED TEXT RECOGNITION MODEL INFERENCE](#CTC-BASED_RECOGNITION)
+    - [3. SRN-BASED TEXT RECOGNITION MODEL INFERENCE](#SRN-BASED_RECOGNITION)
     - [3. TEXT RECOGNITION MODEL INFERENCE USING CUSTOM CHARACTERS DICTIONARY](#USING_CUSTOM_CHARACTERS)
     - [4. MULTILINGUAL MODEL INFERENCE](MULTILINGUAL_MODEL_INFERENCE)
 
@@ -304,8 +305,23 @@ self.character_str = "0123456789abcdefghijklmnopqrstuvwxyz"
 dict_character = list(self.character_str)
 ```
 
+<a name="SRN-BASED_RECOGNITION"></a>
+### 3. SRN-BASED TEXT RECOGNITION MODEL INFERENCE
+
+The recognition model based on SRN requires additional setting of the recognition algorithm parameter
+--rec_algorithm="SRN". At the same time, it is necessary to ensure that the predicted shape is consistent
+with the training, such as: --rec_image_shape="1, 64, 256"
+
+```
+python3 tools/infer/predict_rec.py --image_dir="./doc/imgs_words_en/word_336.png" \
+                                    --rec_model_dir="./inference/srn/" \
+                                    --rec_image_shape="1, 64, 256" \
+                                    --rec_char_type="en" \
+                                    --rec_algorithm="SRN"
+```
+
 <a name="USING_CUSTOM_CHARACTERS"></a>
-### 3. TEXT RECOGNITION MODEL INFERENCE USING CUSTOM CHARACTERS DICTIONARY
+### 4. TEXT RECOGNITION MODEL INFERENCE USING CUSTOM CHARACTERS DICTIONARY
 If the text dictionary is modified during training, when using the inference model to predict, you need to specify the dictionary path used by `--rec_char_dict_path`, and set `rec_char_type=ch`
 
 ```
@@ -313,7 +329,7 @@ python3 tools/infer/predict_rec.py --image_dir="./doc/imgs_words_en/word_336.png
 ```
 
 <a name="MULTILINGUAL_MODEL_INFERENCE"></a>
-### 4. MULTILINGAUL MODEL INFERENCE
+### 5. MULTILINGAUL MODEL INFERENCE
 If you need to predict other language models, when using inference model prediction, you need to specify the dictionary path used by `--rec_char_dict_path`. At the same time, in order to get the correct visualization results,
 You need to specify the visual font path through `--vis_font_path`. There are small language fonts provided by default under the `doc/fonts` path, such as Korean recognition:
 

doc/doc_en/installation_en.md

@@ -3,7 +3,7 @@
 After testing, paddleocr can run on glibc 2.23. You can also test other glibc versions or install glic 2.23 for the best compatibility.
 
 PaddleOCR working environment:
-- PaddlePaddle 1.8+, Recommend PaddlePaddle 2.0rc1
+- PaddlePaddle 2.0.0
 - python3.7
 - glibc 2.23
 
@@ -38,10 +38,10 @@ sudo docker container exec -it ppocr /bin/bash
 pip3 install --upgrade pip
 
 # If you have cuda9 or cuda10 installed on your machine, please run the following command to install
-python3 -m pip install paddlepaddle-gpu==2.0rc1 -i https://mirror.baidu.com/pypi/simple
+python3 -m pip install paddlepaddle-gpu==2.0.0 -i https://mirror.baidu.com/pypi/simple
 
 # If you only have cpu on your machine, please run the following command to install
-python3 -m pip install paddlepaddle==2.0rc1 -i https://mirror.baidu.com/pypi/simple
+python3 -m pip install paddlepaddle==2.0.0 -i https://mirror.baidu.com/pypi/simple
 ```
 For more software version requirements, please refer to the instructions in [Installation Document](https://www.paddlepaddle.org.cn/install/quick) for operation.
 

doc/doc_en/models_list_en.md

@@ -93,7 +93,7 @@ python3 generate_multi_language_configs.py -l it \
 |model name|description|config|model size|download|
 | --- | --- | --- | --- | --- |
 | french_mobile_v2.0_rec |Lightweight model for French recognition|[rec_french_lite_train.yml](../../configs/rec/multi_language/rec_french_lite_train.yml)|2.65M|[inference model](https://paddleocr.bj.bcebos.com/dygraph_v2.0/multilingual/french_mobile_v2.0_rec_infer.tar) / [trained model](https://paddleocr.bj.bcebos.com/dygraph_v2.0/multilingual/french_mobile_v2.0_rec_train.tar) |
-| german_mobile_v2.0_rec |Lightweight model for French recognition|[rec_german_lite_train.yml](../../configs/rec/multi_language/rec_german_lite_train.yml)|2.65M|[inference model](https://paddleocr.bj.bcebos.com/dygraph_v2.0/multilingual/german_mobile_v2.0_rec_infer.tar) / [trained model](https://paddleocr.bj.bcebos.com/dygraph_v2.0/multilingual/german_mobile_v2.0_rec_train.tar) |
+| german_mobile_v2.0_rec |Lightweight model for German recognition|[rec_german_lite_train.yml](../../configs/rec/multi_language/rec_german_lite_train.yml)|2.65M|[inference model](https://paddleocr.bj.bcebos.com/dygraph_v2.0/multilingual/german_mobile_v2.0_rec_infer.tar) / [trained model](https://paddleocr.bj.bcebos.com/dygraph_v2.0/multilingual/german_mobile_v2.0_rec_train.tar) |
 | korean_mobile_v2.0_rec |Lightweight model for Korean recognition|[rec_korean_lite_train.yml](../../configs/rec/multi_language/rec_korean_lite_train.yml)|3.9M|[inference model](https://paddleocr.bj.bcebos.com/dygraph_v2.0/multilingual/korean_mobile_v2.0_rec_infer.tar) / [trained model](https://paddleocr.bj.bcebos.com/dygraph_v2.0/multilingual/korean_mobile_v2.0_rec_train.tar) |
 | japan_mobile_v2.0_rec |Lightweight model for Japanese recognition|[rec_japan_lite_train.yml](../../configs/rec/multi_language/rec_japan_lite_train.yml)|4.23M|[inference model](https://paddleocr.bj.bcebos.com/dygraph_v2.0/multilingual/japan_mobile_v2.0_rec_infer.tar) / [trained model](https://paddleocr.bj.bcebos.com/dygraph_v2.0/multilingual/japan_mobile_v2.0_rec_train.tar) |
 | it_mobile_v2.0_rec |Lightweight model for Italian recognition|rec_it_lite_train.yml|2.53M|[inference model](https://paddleocr.bj.bcebos.com/dygraph_v2.0/multilingual/it_mobile_v2.0_rec_infer.tar) / [trained model](https://paddleocr.bj.bcebos.com/dygraph_v2.0/multilingual/it_mobile_v2.0_rec_train.tar) |

doc/doc_en/recognition_en.md

@@ -126,7 +126,7 @@ In `word_dict.txt`, there is a single word in each line, which maps characters a
 You can use it on demand.
 
 The current multi-language model is still in the demo stage and will continue to optimize the model and add languages. **You are very welcome to provide us with dictionaries and fonts in other languages**,
-If you like, you can submit the dictionary file to [dict](../../ppocr/utils/dict) or corpus file to [corpus](../../ppocr/utils/corpus) and we will thank you in the Repo.
+If you like, you can submit the dictionary file to [dict](../../ppocr/utils/dict) and we will thank you in the Repo.
 
 
 To customize the dict file, please modify the `character_dict_path` field in `configs/rec/rec_icdar15_train.yml` and set `character_type` to `ch`.
@@ -195,6 +195,10 @@ If the evaluation set is large, the test will be time-consuming. It is recommend
 | rec_mv3_none_none_ctc.yml |  Rosetta |   Mobilenet_v3 large 0.5 |  None   |  None |  ctc  |
 | rec_r34_vd_none_bilstm_ctc.yml |  CRNN |   Resnet34_vd |  None   |  BiLSTM |  ctc  |
 | rec_r34_vd_none_none_ctc.yml |  Rosetta |   Resnet34_vd |  None   |  None |  ctc  |
+| rec_mv3_tps_bilstm_att.yml |  CRNN |   Mobilenet_v3 |  TPS   |  BiLSTM |  att  |
+| rec_r34_vd_tps_bilstm_att.yml |  CRNN |   Resnet34_vd |  TPS   |  BiLSTM |  att  |
+| rec_r50fpn_vd_none_srn.yml    | SRN | Resnet50_fpn_vd    | None    | rnn | srn |
+
 
 For training Chinese data, it is recommended to use
 [rec_chinese_lite_train_v2.0.yml](../../configs/rec/ch_ppocr_v2.0/rec_chinese_lite_train_v2.0.yml). If you want to try the result of other algorithms on the Chinese data set, please refer to the following instructions to modify the configuration file:

ppocr/data/__init__.py

@@ -33,7 +33,7 @@ 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 LMDBDateSet
+from ppocr.data.lmdb_dataset import LMDBDataSet
 
 __all__ = ['build_dataloader', 'transform', 'create_operators']
 
@@ -54,7 +54,7 @@ signal.signal(signal.SIGTERM, term_mp)
 def build_dataloader(config, mode, device, logger, seed=None):
     config = copy.deepcopy(config)
 
-    support_dict = ['SimpleDataSet', 'LMDBDateSet']
+    support_dict = ['SimpleDataSet', 'LMDBDataSet']
     module_name = config[mode]['dataset']['name']
     assert module_name in support_dict, Exception(
         'DataSet only support {}'.format(support_dict))

ppocr/data/imaug/__init__.py

@@ -21,7 +21,7 @@ from .make_border_map import MakeBorderMap
 from .make_shrink_map import MakeShrinkMap
 from .random_crop_data import EastRandomCropData, PSERandomCrop
 
-from .rec_img_aug import RecAug, RecResizeImg, ClsResizeImg
+from .rec_img_aug import RecAug, RecResizeImg, ClsResizeImg, SRNRecResizeImg
 from .randaugment import RandAugment
 from .operators import *
 from .label_ops import *

ppocr/data/imaug/label_ops.py

@@ -102,6 +102,8 @@ class BaseRecLabelEncode(object):
             support_character_type, character_type)
 
         self.max_text_len = max_text_length
+        self.beg_str = "sos"
+        self.end_str = "eos"
         if character_type == "en":
             self.character_str = "0123456789abcdefghijklmnopqrstuvwxyz"
             dict_character = list(self.character_str)
@@ -197,16 +199,76 @@ class AttnLabelEncode(BaseRecLabelEncode):
         super(AttnLabelEncode,
               self).__init__(max_text_length, character_dict_path,
                              character_type, use_space_char)
+
+    def add_special_char(self, dict_character):
         self.beg_str = "sos"
         self.end_str = "eos"
+        dict_character = [self.beg_str] + dict_character + [self.end_str]
+        return dict_character
+
+    def __call__(self, data):
+        text = data['label']
+        text = self.encode(text)
+        if text is None:
+            return None
+        if len(text) >= self.max_text_len:
+            return None
+        data['length'] = np.array(len(text))
+        text = [0] + text + [len(self.character) - 1] + [0] * (self.max_text_len
+                                                               - len(text) - 1)
+        data['label'] = np.array(text)
+        return data
+
+    def get_ignored_tokens(self):
+        beg_idx = self.get_beg_end_flag_idx("beg")
+        end_idx = self.get_beg_end_flag_idx("end")
+        return [beg_idx, end_idx]
+
+    def get_beg_end_flag_idx(self, beg_or_end):
+        if beg_or_end == "beg":
+            idx = np.array(self.dict[self.beg_str])
+        elif beg_or_end == "end":
+            idx = np.array(self.dict[self.end_str])
+        else:
+            assert False, "Unsupport type %s in get_beg_end_flag_idx" \
+                          % beg_or_end
+        return idx
+
+
+class SRNLabelEncode(BaseRecLabelEncode):
+    """ Convert between text-label and text-index """
+
+    def __init__(self,
+                 max_text_length=25,
+                 character_dict_path=None,
+                 character_type='en',
+                 use_space_char=False,
+                 **kwargs):
+        super(SRNLabelEncode,
+              self).__init__(max_text_length, character_dict_path,
+                             character_type, use_space_char)
 
     def add_special_char(self, dict_character):
-        dict_character = [self.beg_str, self.end_str] + dict_character
+        dict_character = dict_character + [self.beg_str, self.end_str]
         return dict_character
 
-    def __call__(self, text):
+    def __call__(self, data):
+        text = data['label']
         text = self.encode(text)
-        return text
+        char_num = len(self.character_str)
+        if text is None:
+            return None
+        if len(text) > self.max_text_len:
+            return None
+        data['length'] = np.array(len(text))
+        text = text + [char_num] * (self.max_text_len - len(text))
+        data['label'] = np.array(text)
+        return data
+
+    def get_ignored_tokens(self):
+        beg_idx = self.get_beg_end_flag_idx("beg")
+        end_idx = self.get_beg_end_flag_idx("end")
+        return [beg_idx, end_idx]
 
     def get_beg_end_flag_idx(self, beg_or_end):
         if beg_or_end == "beg":

ppocr/data/imaug/rec_img_aug.py

@@ -12,20 +12,6 @@
 # See the License for the specific language governing permissions and
 # limitations under the License.
 
-# copyright (c) 2020 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 math
 import cv2
 import numpy as np
@@ -77,6 +63,26 @@ class RecResizeImg(object):
         return data
 
 
+class SRNRecResizeImg(object):
+    def __init__(self, image_shape, num_heads, max_text_length, **kwargs):
+        self.image_shape = image_shape
+        self.num_heads = num_heads
+        self.max_text_length = max_text_length
+
+    def __call__(self, data):
+        img = data['image']
+        norm_img = resize_norm_img_srn(img, self.image_shape)
+        data['image'] = norm_img
+        [encoder_word_pos, gsrm_word_pos, gsrm_slf_attn_bias1, gsrm_slf_attn_bias2] = \
+            srn_other_inputs(self.image_shape, self.num_heads, self.max_text_length)
+
+        data['encoder_word_pos'] = encoder_word_pos
+        data['gsrm_word_pos'] = gsrm_word_pos
+        data['gsrm_slf_attn_bias1'] = gsrm_slf_attn_bias1
+        data['gsrm_slf_attn_bias2'] = gsrm_slf_attn_bias2
+        return data
+
+
 def resize_norm_img(img, image_shape):
     imgC, imgH, imgW = image_shape
     h = img.shape[0]
@@ -103,7 +109,7 @@ def resize_norm_img(img, image_shape):
 def resize_norm_img_chinese(img, image_shape):
     imgC, imgH, imgW = image_shape
     # todo: change to 0 and modified image shape
-    max_wh_ratio = 0
+    max_wh_ratio = imgW * 1.0 / imgH
     h, w = img.shape[0], img.shape[1]
     ratio = w * 1.0 / h
     max_wh_ratio = max(max_wh_ratio, ratio)
@@ -126,6 +132,60 @@ def resize_norm_img_chinese(img, image_shape):
     return padding_im
 
 
+def resize_norm_img_srn(img, image_shape):
+    imgC, imgH, imgW = image_shape
+
+    img_black = np.zeros((imgH, imgW))
+    im_hei = img.shape[0]
+    im_wid = img.shape[1]
+
+    if im_wid <= im_hei * 1:
+        img_new = cv2.resize(img, (imgH * 1, imgH))
+    elif im_wid <= im_hei * 2:
+        img_new = cv2.resize(img, (imgH * 2, imgH))
+    elif im_wid <= im_hei * 3:
+        img_new = cv2.resize(img, (imgH * 3, imgH))
+    else:
+        img_new = cv2.resize(img, (imgW, imgH))
+
+    img_np = np.asarray(img_new)
+    img_np = cv2.cvtColor(img_np, cv2.COLOR_BGR2GRAY)
+    img_black[:, 0:img_np.shape[1]] = img_np
+    img_black = img_black[:, :, np.newaxis]
+
+    row, col, c = img_black.shape
+    c = 1
+
+    return np.reshape(img_black, (c, row, col)).astype(np.float32)
+
+
+def srn_other_inputs(image_shape, num_heads, max_text_length):
+
+    imgC, imgH, imgW = image_shape
+    feature_dim = int((imgH / 8) * (imgW / 8))
+
+    encoder_word_pos = np.array(range(0, feature_dim)).reshape(
+        (feature_dim, 1)).astype('int64')
+    gsrm_word_pos = np.array(range(0, max_text_length)).reshape(
+        (max_text_length, 1)).astype('int64')
+
+    gsrm_attn_bias_data = np.ones((1, max_text_length, max_text_length))
+    gsrm_slf_attn_bias1 = np.triu(gsrm_attn_bias_data, 1).reshape(
+        [1, max_text_length, max_text_length])
+    gsrm_slf_attn_bias1 = np.tile(gsrm_slf_attn_bias1,
+                                  [num_heads, 1, 1]) * [-1e9]
+
+    gsrm_slf_attn_bias2 = np.tril(gsrm_attn_bias_data, -1).reshape(
+        [1, max_text_length, max_text_length])
+    gsrm_slf_attn_bias2 = np.tile(gsrm_slf_attn_bias2,
+                                  [num_heads, 1, 1]) * [-1e9]
+
+    return [
+        encoder_word_pos, gsrm_word_pos, gsrm_slf_attn_bias1,
+        gsrm_slf_attn_bias2
+    ]
+
+
 def flag():
     """
     flag

ppocr/data/lmdb_dataset.py

@@ -20,9 +20,9 @@ import cv2
 from .imaug import transform, create_operators
 
 
-class LMDBDateSet(Dataset):
+class LMDBDataSet(Dataset):
     def __init__(self, config, mode, logger, seed=None):
-        super(LMDBDateSet, self).__init__()
+        super(LMDBDataSet, self).__init__()
 
         global_config = config['Global']
         dataset_config = config[mode]['dataset']

ppocr/losses/__init__.py

@@ -23,11 +23,16 @@ def build_loss(config):
 
     # rec loss
     from .rec_ctc_loss import CTCLoss
+    from .rec_att_loss import AttentionLoss
+    from .rec_srn_loss import SRNLoss
 
     # cls loss
     from .cls_loss import ClsLoss
 
-    support_dict = ['DBLoss', 'EASTLoss', 'SASTLoss', 'CTCLoss', 'ClsLoss']
+    support_dict = [
+        'DBLoss', 'EASTLoss', 'SASTLoss', 'CTCLoss', 'ClsLoss', 'AttentionLoss',
+        'SRNLoss'
+    ]
 
     config = copy.deepcopy(config)
     module_name = config.pop('name')

ppocr/losses/rec_att_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
+
+import paddle
+from paddle import nn
+
+
+class AttentionLoss(nn.Layer):
+    def __init__(self, **kwargs):
+        super(AttentionLoss, self).__init__()
+        self.loss_func = nn.CrossEntropyLoss(weight=None, reduction='none')
+
+    def forward(self, predicts, batch):
+        targets = batch[1].astype("int64")
+        label_lengths = batch[2].astype('int64')
+        batch_size, num_steps, num_classes = predicts.shape[0], predicts.shape[
+            1], predicts.shape[2]
+        assert len(targets.shape) == len(list(predicts.shape)) - 1, \
+            "The target's shape and inputs's shape is [N, d] and [N, num_steps]"
+
+        inputs = paddle.reshape(predicts, [-1, predicts.shape[-1]])
+        targets = paddle.reshape(targets, [-1])
+
+        return {'loss': paddle.sum(self.loss_func(inputs, targets))}

ppocr/losses/rec_srn_loss.py

+# copyright (c) 2020 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
+
+
+class SRNLoss(nn.Layer):
+    def __init__(self, **kwargs):
+        super(SRNLoss, self).__init__()
+        self.loss_func = paddle.nn.loss.CrossEntropyLoss(reduction="sum")
+
+    def forward(self, predicts, batch):
+        predict = predicts['predict']
+        word_predict = predicts['word_out']
+        gsrm_predict = predicts['gsrm_out']
+        label = batch[1]
+
+        casted_label = paddle.cast(x=label, dtype='int64')
+        casted_label = paddle.reshape(x=casted_label, shape=[-1, 1])
+
+        cost_word = self.loss_func(word_predict, label=casted_label)
+        cost_gsrm = self.loss_func(gsrm_predict, label=casted_label)
+        cost_vsfd = self.loss_func(predict, label=casted_label)
+
+        cost_word = paddle.reshape(x=paddle.sum(cost_word), shape=[1])
+        cost_gsrm = paddle.reshape(x=paddle.sum(cost_gsrm), shape=[1])
+        cost_vsfd = paddle.reshape(x=paddle.sum(cost_vsfd), shape=[1])
+
+        sum_cost = cost_word * 3.0 + cost_vsfd + cost_gsrm * 0.15
+
+        return {'loss': sum_cost, 'word_loss': cost_word, 'img_loss': cost_vsfd}

ppocr/metrics/rec_metric.py

@@ -33,8 +33,6 @@ class RecMetric(object):
             if pred == target:
                 correct_num += 1
             all_num += 1
-            # if all_num < 10 and kwargs.get('show_str', False):
-            #     print('{} -> {}'.format(pred, target))
         self.correct_num += correct_num
         self.all_num += all_num
         self.norm_edit_dis += norm_edit_dis
@@ -50,7 +48,7 @@ class RecMetric(object):
                  'norm_edit_dis': 0,
             }
         """
-        acc = self.correct_num / self.all_num
+        acc = 1.0 * self.correct_num / self.all_num
         norm_edit_dis = 1 - self.norm_edit_dis / self.all_num
         self.reset()
         return {'acc': acc, 'norm_edit_dis': norm_edit_dis}

ppocr/modeling/architectures/base_model.py

@@ -68,11 +68,14 @@ class BaseModel(nn.Layer):
         config["Head"]['in_channels'] = in_channels
         self.head = build_head(config["Head"])
 
-    def forward(self, x):
+    def forward(self, x, data=None):
         if self.use_transform:
             x = self.transform(x)
         x = self.backbone(x)
         if self.use_neck:
             x = self.neck(x)
+        if data is None:
             x = self.head(x)
+        else:
+            x = self.head(x, data)
         return x

ppocr/modeling/backbones/__init__.py

@@ -24,7 +24,8 @@ def build_backbone(config, model_type):
     elif model_type == 'rec' or model_type == 'cls':
         from .rec_mobilenet_v3 import MobileNetV3
         from .rec_resnet_vd import ResNet
-        support_dict = ['MobileNetV3', 'ResNet', 'ResNet_FPN']
+        from .rec_resnet_fpn import ResNetFPN
+        support_dict = ['MobileNetV3', 'ResNet', 'ResNetFPN']
     else:
         raise NotImplementedError