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

2021-10-08 11:50:39 +08:00 · 2021-10-08 11:50:39 +08:00 · 17e1e2c5ad
parent cf35373ba1 f14b79b489
commit 17e1e2c5ad
37 changed files with 1755 additions and 75 deletions
--- a/benchmark/run_benchmark_det.sh
+++ b/benchmark/run_benchmark_det.sh
@ -0,0 +1,54 @@
+#!/usr/bin/env bash
+set -xe
+# 运行示例：CUDA_VISIBLE_DEVICES=0 bash run_benchmark.sh ${run_mode} ${bs_item} ${fp_item} 500 ${model_mode}
+# 参数说明
+function _set_params(){
+    run_mode=${1:-"sp"}          # 单卡sp|多卡mp
+    batch_size=${2:-"64"}
+    fp_item=${3:-"fp32"}        # fp32|fp16
+    max_iter=${4:-"500"}       # 可选，如果需要修改代码提前中断
+    model_name=${5:-"model_name"}
+    run_log_path=${TRAIN_LOG_DIR:-$(pwd)}  # TRAIN_LOG_DIR 后续QA设置该参数
+
+#   以下不用修改   
+    device=${CUDA_VISIBLE_DEVICES//,/ }
+    arr=(${device})
+    num_gpu_devices=${#arr[*]}
+    log_file=${run_log_path}/${model_name}_${run_mode}_bs${batch_size}_${fp_item}_${num_gpu_devices}
+}
+function _train(){
+    echo "Train on ${num_gpu_devices} GPUs"
+    echo "current CUDA_VISIBLE_DEVICES=$CUDA_VISIBLE_DEVICES, gpus=$num_gpu_devices, batch_size=$batch_size"
+
+    train_cmd="-c configs/det/${model_name}.yml
+               -o Train.loader.batch_size_per_card=${batch_size}
+               -o Global.epoch_num=${max_iter} "   
+    case ${run_mode} in
+      sp) 
+        train_cmd="python3.7 tools/train.py "${train_cmd}""
+        ;;
+      mp)
+        train_cmd="python3.7 -m paddle.distributed.launch --log_dir=./mylog --gpus=$CUDA_VISIBLE_DEVICES tools/train.py ${train_cmd}"
+        ;;
+      *) echo "choose run_mode(sp or mp)"; exit 1;
+    esac
+# 以下不用修改
+    timeout 15m ${train_cmd} > ${log_file} 2>&1
+    if [ $? -ne 0 ];then
+            echo -e "${model_name}, FAIL"
+        export job_fail_flag=1
+    else
+        echo -e "${model_name}, SUCCESS"
+        export job_fail_flag=0
+    fi
+    kill -9 `ps -ef|grep 'python3.7'|awk '{print $2}'`
+
+    if [ $run_mode = "mp" -a -d mylog ]; then
+        rm ${log_file}
+        cp mylog/workerlog.0 ${log_file}
+    fi
+}
+
+_set_params $@
+_train
+
--- a/benchmark/run_det.sh
+++ b/benchmark/run_det.sh
@ -0,0 +1,29 @@
+# 提供可稳定复现性能的脚本，默认在标准docker环境内py37执行： paddlepaddle/paddle:latest-gpu-cuda10.1-cudnn7  paddle=2.1.2  py=37
+# 执行目录：需说明
+cd PaddleOCR
+# 1 安装该模型需要的依赖 (如需开启优化策略请注明)
+python3.7 -m pip install -r requirements.txt
+# 2 拷贝该模型需要数据、预训练模型
+wget -p ./tain_data/  https://paddleocr.bj.bcebos.com/dygraph_v2.0/test/icdar2015.tar && cd train_data  && tar xf icdar2015.tar && cd ../
+wget -p ./pretrain_models/ https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/ResNet50_vd_pretrained.pdparams
+# 3 批量运行（如不方便批量，1，2需放到单个模型中）
+
+model_mode_list=(det_mv3_db det_r50_vd_east)
+fp_item_list=(fp32)
+bs_list=(256 128)
+for model_mode in ${model_mode_list[@]}; do
+      for fp_item in ${fp_item_list[@]}; do
+          for bs_item in ${bs_list[@]}; do
+            echo "index is speed, 1gpus, begin, ${model_name}"
+            run_mode=sp
+            CUDA_VISIBLE_DEVICES=0 bash benchmark/run_benchmark.sh ${run_mode} ${bs_item} ${fp_item} 10 ${model_mode}     #  (5min)
+            sleep 60
+            echo "index is speed, 8gpus, run_mode is multi_process, begin, ${model_name}"
+            run_mode=mp
+            CUDA_VISIBLE_DEVICES=0,1,2,3,4,5,6,7 bash benchmark/run_benchmark.sh ${run_mode} ${bs_item} ${fp_item} 10 ${model_mode} 
+            sleep 60
+            done
+      done
+done
+
+
--- a/configs/det/det_r50_vd_east.yml
+++ b/configs/det/det_r50_vd_east.yml
@ -8,7 +8,7 @@ Global:
  # evaluation is run every 5000 iterations after the 4000th iteration
  eval_batch_step: [4000, 5000]
  cal_metric_during_train: False
-  pretrained_model: ./pretrain_models/ResNet50_vd_pretrained/
+  pretrained_model: ./pretrain_models/ResNet50_vd_pretrained
  checkpoints: 
  save_inference_dir:
  use_visualdl: False
--- a/configs/rec/ch_PP-OCRv2/ch_PP-OCRv2_rec_enhanced_ctc_loss.yml
+++ b/configs/rec/ch_PP-OCRv2/ch_PP-OCRv2_rec_enhanced_ctc_loss.yml
@ -0,0 +1,126 @@
+Global:
+  debug: false
+  use_gpu: true
+  epoch_num: 800
+  log_smooth_window: 20
+  print_batch_step: 10
+  save_model_dir: ./output/rec_mobile_pp-OCRv2_enhanced_ctc_loss
+  save_epoch_step: 3
+  eval_batch_step: [0, 2000]
+  cal_metric_during_train: true
+  pretrained_model:
+  checkpoints:
+  save_inference_dir:
+  use_visualdl: false
+  infer_img: doc/imgs_words/ch/word_1.jpg
+  character_dict_path: ppocr/utils/ppocr_keys_v1.txt
+  character_type: ch
+  max_text_length: 25
+  infer_mode: false
+  use_space_char: true
+  distributed: true
+  save_res_path: ./output/rec/predicts_mobile_pp-OCRv2_enhanced_ctc_loss.txt
+
+
+Optimizer:
+  name: Adam
+  beta1: 0.9
+  beta2: 0.999
+  lr:
+    name: Piecewise
+    decay_epochs : [700, 800]
+    values : [0.001, 0.0001]
+    warmup_epoch: 5
+  regularizer:
+    name: L2
+    factor: 2.0e-05
+
+
+Architecture:
+  model_type: rec
+  algorithm: CRNN
+  Transform:
+  Backbone:
+    name: MobileNetV1Enhance
+    scale: 0.5
+  Neck:
+    name: SequenceEncoder
+    encoder_type: rnn
+    hidden_size: 64
+  Head:
+    name: CTCHead
+    mid_channels: 96
+    fc_decay: 0.00002
+    return_feats: true
+
+Loss:
+  name: CombinedLoss
+  loss_config_list:
+  - CTCLoss:
+      use_focal_loss: false
+      weight: 1.0
+  - CenterLoss:
+      weight: 0.05
+      num_classes: 6625
+      feat_dim: 96
+      init_center: false
+      center_file_path: "./train_center.pkl"
+  # you can also try to add ace loss on your own dataset
+  # - ACELoss:
+  #     weight: 0.1
+
+PostProcess:
+  name: CTCLabelDecode
+
+Metric:
+  name: RecMetric
+  main_indicator: acc
+
+Train:
+  dataset:
+    name: SimpleDataSet
+    data_dir: ./train_data/
+    label_file_list:
+    - ./train_data/train_list.txt
+    transforms:
+    - DecodeImage:
+        img_mode: BGR
+        channel_first: false
+    - RecAug:
+    - CTCLabelEncode:
+    - RecResizeImg:
+        image_shape: [3, 32, 320]
+    - KeepKeys:
+        keep_keys:
+        - image
+        - label
+        - length
+        - label_ace
+  loader:
+    shuffle: true
+    batch_size_per_card: 128
+    drop_last: true
+    num_workers: 8
+Eval:
+  dataset:
+    name: SimpleDataSet
+    data_dir: ./train_data
+    label_file_list:
+    - ./train_data/val_list.txt
+    transforms:
+    - DecodeImage:
+        img_mode: BGR
+        channel_first: false
+    - CTCLabelEncode:
+    - RecResizeImg:
+        image_shape: [3, 32, 320]
+    - KeepKeys:
+        keep_keys:
+        - image
+        - label
+        - length
+  loader:
+    shuffle: false
+    drop_last: false
+    batch_size_per_card: 128
+    num_workers: 8
--- a/configs/rec/rec_resnet_stn_bilstm_att.yml
+++ b/configs/rec/rec_resnet_stn_bilstm_att.yml
@ -0,0 +1,109 @@
+Global:
+  use_gpu: True
+  epoch_num: 400
+  log_smooth_window: 20
+  print_batch_step: 10
+  save_model_dir: ./output/rec/seed
+  save_epoch_step: 3
+  # evaluation is run every 5000 iterations after the 4000th iteration
+  eval_batch_step: [0, 2000]
+  cal_metric_during_train: True
+  pretrained_model:
+  checkpoints:
+  save_inference_dir:
+  use_visualdl: False
+  infer_img: doc/imgs_words_en/word_10.png
+  # for data or label process
+  character_dict_path: 
+  character_type: EN_symbol
+  max_text_length: 100
+  infer_mode: False
+  use_space_char: False
+  save_res_path: ./output/rec/predicts_seed.txt
+
+
+Optimizer:
+  name: Adadelta
+  weight_deacy: 0.0
+  momentum: 0.9
+  lr:
+    name: Piecewise
+    decay_epochs: [4,5,8]
+    values: [1.0, 0.1, 0.01]
+  regularizer:
+    name: 'L2'
+    factor: 2.0e-05
+
+
+Architecture:
+  model_type: rec
+  algorithm: SEED
+  Transform:
+    name: STN_ON
+    tps_inputsize: [32, 64]
+    tps_outputsize: [32, 100]
+    num_control_points: 20
+    tps_margins: [0.05,0.05]
+    stn_activation: none
+  Backbone:
+    name: ResNet_ASTER
+  Head:
+    name: AsterHead  # AttentionHead
+    sDim: 512
+    attDim: 512
+    max_len_labels: 100
+
+Loss:
+  name: AsterLoss
+
+PostProcess:
+  name: SEEDLabelDecode
+
+Metric:
+  name: RecMetric
+  main_indicator: acc
+  is_filter: True
+
+Train:
+  dataset:
+    name: LMDBDataSet
+    data_dir: ./train_data/data_lmdb_release/training/
+    transforms:
+      - Fasttext:
+          path: "./cc.en.300.bin"
+      - DecodeImage: # load image
+          img_mode: BGR
+          channel_first: False
+      - SEEDLabelEncode: # Class handling label
+      - RecResizeImg:
+          character_type: en
+          image_shape: [3, 64, 256]
+          padding: False
+      - KeepKeys:
+          keep_keys: ['image', 'label', 'length', 'fast_label'] # dataloader will return list in this order
+  loader:
+    shuffle: True
+    batch_size_per_card: 256
+    drop_last: True
+    num_workers: 6
+
+Eval:
+  dataset:
+    name: LMDBDataSet
+    data_dir: ./train_data/data_lmdb_release/evaluation/
+    transforms:
+      - DecodeImage: # load image
+          img_mode: BGR
+          channel_first: False
+      - SEEDLabelEncode: # Class handling label
+      - RecResizeImg:
+          character_type: en
+          image_shape: [3, 64, 256]
+          padding: False
+      - KeepKeys:
+          keep_keys: ['image', 'label', 'length'] # dataloader will return list in this order
+  loader:
+    shuffle: False
+    drop_last: True
+    batch_size_per_card: 256
+    num_workers: 4
--- a/deploy/slim/prune/sensitivity_anal.py
+++ b/deploy/slim/prune/sensitivity_anal.py
@ -110,25 +110,42 @@ def main(config, device, logger, vdl_writer):
        logger.info("metric['hmean']: {}".format(metric['hmean']))
        return metric['hmean']

-    params_sensitive = pruner.sensitive(
-        eval_func=eval_fn,
-        sen_file="./sen.pickle",
-        skip_vars=[
-            "conv2d_57.w_0", "conv2d_transpose_2.w_0", "conv2d_transpose_3.w_0"
-        ])
+    run_sensitive_analysis = False
+    """
+    run_sensitive_analysis=True: 
+        Automatically compute the sensitivities of convolutions in a model. 
+        The sensitivity of a convolution is the losses of accuracy on test dataset in 
+        differenct pruned ratios. The sensitivities can be used to get a group of best 
+        ratios with some condition.
+    
+    run_sensitive_analysis=False: 
+        Set prune trim ratio to a fixed value, such as 10%. The larger the value, 
+        the more convolution weights will be cropped.

-    logger.info(
-        "The sensitivity analysis results of model parameters saved in sen.pickle"
-    )
-    # calculate pruned params's ratio
-    params_sensitive = pruner._get_ratios_by_loss(params_sensitive, loss=0.02)
-    for key in params_sensitive.keys():
-        logger.info("{}, {}".format(key, params_sensitive[key]))
+    """

-    #params_sensitive = {}
-    #for param in model.parameters():
-    #    if 'transpose' not in param.name and 'linear' not in param.name:
-    #        params_sensitive[param.name] = 0.1  
+    if run_sensitive_analysis:
+        params_sensitive = pruner.sensitive(
+            eval_func=eval_fn,
+            sen_file="./deploy/slim/prune/sen.pickle",
+            skip_vars=[
+                "conv2d_57.w_0", "conv2d_transpose_2.w_0",
+                "conv2d_transpose_3.w_0"
+            ])
+        logger.info(
+            "The sensitivity analysis results of model parameters saved in sen.pickle"
+        )
+        # calculate pruned params's ratio
+        params_sensitive = pruner._get_ratios_by_loss(
+            params_sensitive, loss=0.02)
+        for key in params_sensitive.keys():
+            logger.info("{}, {}".format(key, params_sensitive[key]))
+    else:
+        params_sensitive = {}
+        for param in model.parameters():
+            if 'transpose' not in param.name and 'linear' not in param.name:
+                # set prune ratio as 10%. The larger the value, the more convolution weights will be cropped
+                params_sensitive[param.name] = 0.1

    plan = pruner.prune_vars(params_sensitive, [0])

--- a/doc/doc_ch/algorithm_overview.md
+++ b/doc/doc_ch/algorithm_overview.md
@ -50,6 +50,7 @@ PaddleOCR基于动态图开源的文本识别算法列表：
 - [x]  SRN([paper](https://arxiv.org/abs/2003.12294))
 - [x]  NRTR([paper](https://arxiv.org/abs/1806.00926v2))
 - [x]  SAR([paper](https://arxiv.org/abs/1811.00751v2))
+- [x]  SEED([paper](https://arxiv.org/pdf/2005.10977.pdf))

 参考[DTRB](https://arxiv.org/abs/1904.01906) 文字识别训练和评估流程，使用MJSynth和SynthText两个文字识别数据集训练，在IIIT, SVT, IC03, IC13, IC15, SVTP, CUTE数据集上进行评估，算法效果如下：

@ -66,5 +67,5 @@ PaddleOCR基于动态图开源的文本识别算法列表：
 |SRN|Resnet50_vd_fpn| 88.52% | rec_r50fpn_vd_none_srn | [下载链接](https://paddleocr.bj.bcebos.com/dygraph_v2.0/en/rec_r50_vd_srn_train.tar) |
 |NRTR|NRTR_MTB| 84.3% | rec_mtb_nrtr | [下载链接](https://paddleocr.bj.bcebos.com/dygraph_v2.0/en/rec_mtb_nrtr_train.tar) |
 |SAR|Resnet31| 87.2% | rec_r31_sar | [下载链接](https://paddleocr.bj.bcebos.com/dygraph_v2.1/rec/rec_r31_sar_train.tar) |
-
+|SEED| Aster_Resnet | 85.2% | rec_resnet_stn_bilstm_att | [下载链接](https://paddleocr.bj.bcebos.com/dygraph_v2.1/rec/rec_resnet_stn_bilstm_att.tar)|
 PaddleOCR文本识别算法的训练和使用请参考文档教程中[模型训练/评估中的文本识别部分](./recognition.md)。
--- a/doc/doc_ch/recognition.md
+++ b/doc/doc_ch/recognition.md
@ -234,6 +234,9 @@ PaddleOCR支持训练和评估交替进行, 可以在 `configs/rec/rec_icdar15_t
 | rec_r50fpn_vd_none_srn.yml    | SRN | Resnet50_fpn_vd    | None    | rnn | srn |
 | rec_mtb_nrtr.yml    | NRTR | nrtr_mtb    | None    | transformer encoder | transformer decoder |
 | rec_r31_sar.yml               | SAR | ResNet31 | None | LSTM encoder | LSTM decoder |
+| rec_resnet_stn_bilstm_att.yml | SEED | Aster_Resnet | STN | BiLSTM | att |
+
+*其中SEED模型需要额外加载FastText训练好的[语言模型](https://dl.fbaipublicfiles.com/fasttext/vectors-crawl/cc.en.300.bin.gz)

 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：

@ -460,5 +463,3 @@ python3 tools/export_model.py -c configs/rec/ch_ppocr_v2.0/rec_chinese_lite_trai
  ```
  python3 tools/infer/predict_rec.py --image_dir="./doc/imgs_words_en/word_336.png" --rec_model_dir="./your inference model" --rec_image_shape="3, 32, 100" --rec_char_type="ch" --rec_char_dict_path="your text dict path"
  ```
-
-  
--- a/doc/joinus.PNG
+++ b/doc/joinus.PNG
--- a/ppocr/data/imaug/label_ops.py
+++ b/ppocr/data/imaug/label_ops.py
@ -215,6 +215,11 @@ class CTCLabelEncode(BaseRecLabelEncode):
        data['length'] = np.array(len(text))
        text = text + [0] * (self.max_text_len - len(text))
        data['label'] = np.array(text)
+
+        label = [0] * len(self.character)
+        for x in text:
+            label[x] += 1
+        data['label_ace'] = np.array(label)
        return data

    def add_special_char(self, dict_character):
@ -342,6 +347,38 @@ class AttnLabelEncode(BaseRecLabelEncode):
        return idx


+class SEEDLabelEncode(BaseRecLabelEncode):
+    """ Convert between text-label and text-index """
+
+    def __init__(self,
+                 max_text_length,
+                 character_dict_path=None,
+                 character_type='ch',
+                 use_space_char=False,
+                 **kwargs):
+        super(SEEDLabelEncode,
+              self).__init__(max_text_length, character_dict_path,
+                             character_type, use_space_char)
+
+    def add_special_char(self, dict_character):
+        self.end_str = "eos"
+        dict_character = 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)) + 1  # conclude eos
+        text = text + [len(self.character) - 1] * (self.max_text_len - len(text)
+                                                   )
+        data['label'] = np.array(text)
+        return data
+
+
 class SRNLabelEncode(BaseRecLabelEncode):
    """ Convert between text-label and text-index """

@ -421,7 +458,6 @@ class TableLabelEncode(object):
            substr = lines[0].decode('utf-8').strip("\r\n").split("\t")
            character_num = int(substr[0])
            elem_num = int(substr[1])
-
            for cno in range(1, 1 + character_num):
                character = lines[cno].decode('utf-8').strip("\r\n")
                list_character.append(character)
--- a/ppocr/data/imaug/operators.py
+++ b/ppocr/data/imaug/operators.py
@ -23,6 +23,7 @@ import sys
 import six
 import cv2
 import numpy as np
+import fasttext


 class DecodeImage(object):
@ -83,12 +84,13 @@ class NRTRDecodeImage(object):
        elif self.img_mode == 'RGB':
            assert img.shape[2] == 3, 'invalid shape of image[%s]' % (img.shape)
            img = img[:, :, ::-1]
-        img = cv2.cvtColor(img,cv2.COLOR_BGR2GRAY)
+        img = cv2.cvtColor(img, cv2.COLOR_BGR2GRAY)
        if self.channel_first:
            img = img.transpose((2, 0, 1))
        data['image'] = img
        return data

+
 class NormalizeImage(object):
    """ normalize image such as substract mean, divide std
    """
@ -133,6 +135,17 @@ class ToCHWImage(object):
        return data


+class Fasttext(object):
+    def __init__(self, path="None", **kwargs):
+        self.fast_model = fasttext.load_model(path)
+
+    def __call__(self, data):
+        label = data['label']
+        fast_label = self.fast_model[label]
+        data['fast_label'] = fast_label
+        return data
+
+
 class KeepKeys(object):
    def __init__(self, keep_keys, **kwargs):
        self.keep_keys = keep_keys
--- a/ppocr/data/imaug/rec_img_aug.py
+++ b/ppocr/data/imaug/rec_img_aug.py
@ -88,17 +88,19 @@ class RecResizeImg(object):
                 image_shape,
                 infer_mode=False,
                 character_type='ch',
+                 padding=True,
                 **kwargs):
        self.image_shape = image_shape
        self.infer_mode = infer_mode
        self.character_type = character_type
+        self.padding = padding

    def __call__(self, data):
        img = data['image']
        if self.infer_mode and self.character_type == "ch":
            norm_img = resize_norm_img_chinese(img, self.image_shape)
        else:
-            norm_img = resize_norm_img(img, self.image_shape)
+            norm_img = resize_norm_img(img, self.image_shape, self.padding)
        data['image'] = norm_img
        return data

@ -174,16 +176,21 @@ def resize_norm_img_sar(img, image_shape, width_downsample_ratio=0.25):
    return padding_im, resize_shape, pad_shape, valid_ratio


-def resize_norm_img(img, image_shape):
+def resize_norm_img(img, image_shape, padding=True):
    imgC, imgH, imgW = image_shape
    h = img.shape[0]
    w = img.shape[1]
-    ratio = w / float(h)
-    if math.ceil(imgH * ratio) > imgW:
+    if not padding:
+        resized_image = cv2.resize(
+            img, (imgW, imgH), interpolation=cv2.INTER_LINEAR)
        resized_w = imgW
    else:
-        resized_w = int(math.ceil(imgH * ratio))
-    resized_image = cv2.resize(img, (resized_w, imgH))
+        ratio = w / float(h)
+        if math.ceil(imgH * ratio) > imgW:
+            resized_w = imgW
+        else:
+            resized_w = int(math.ceil(imgH * ratio))
+        resized_image = cv2.resize(img, (resized_w, imgH))
    resized_image = resized_image.astype('float32')
    if image_shape[0] == 1:
        resized_image = resized_image / 255
--- a/ppocr/losses/init.py
+++ b/ppocr/losses/init.py
@ -28,6 +28,8 @@ from .rec_att_loss import AttentionLoss
 from .rec_srn_loss import SRNLoss
 from .rec_nrtr_loss import NRTRLoss
 from .rec_sar_loss import SARLoss
+from .rec_aster_loss import AsterLoss
+
 # cls loss
 from .cls_loss import ClsLoss

@ -48,9 +50,8 @@ def build_loss(config):
    support_dict = [
        'DBLoss', 'PSELoss', 'EASTLoss', 'SASTLoss', 'CTCLoss', 'ClsLoss',
        'AttentionLoss', 'SRNLoss', 'PGLoss', 'CombinedLoss', 'NRTRLoss',
-        'TableAttentionLoss', 'SARLoss'
+        'TableAttentionLoss', 'SARLoss', 'AsterLoss'
    ]
-
    config = copy.deepcopy(config)
    module_name = config.pop('name')
    assert module_name in support_dict, Exception('loss only support {}'.format(
--- a/ppocr/losses/ace_loss.py
+++ b/ppocr/losses/ace_loss.py
@ -0,0 +1,50 @@
+# 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
+import paddle.nn as nn
+
+
+class ACELoss(nn.Layer):
+    def __init__(self, **kwargs):
+        super().__init__()
+        self.loss_func = nn.CrossEntropyLoss(
+            weight=None,
+            ignore_index=0,
+            reduction='none',
+            soft_label=True,
+            axis=-1)
+
+    def __call__(self, predicts, batch):
+        if isinstance(predicts, (list, tuple)):
+            predicts = predicts[-1]
+        B, N = predicts.shape[:2]
+        div = paddle.to_tensor([N]).astype('float32')
+
+        predicts = nn.functional.softmax(predicts, axis=-1)
+        aggregation_preds = paddle.sum(predicts, axis=1)
+        aggregation_preds = paddle.divide(aggregation_preds, div)
+
+        length = batch[2].astype("float32")
+        batch = batch[3].astype("float32")
+        batch[:, 0] = paddle.subtract(div, length)
+
+        batch = paddle.divide(batch, div)
+
+        loss = self.loss_func(aggregation_preds, batch)
+
+        return {"loss_ace": loss}
--- a/ppocr/losses/center_loss.py
+++ b/ppocr/losses/center_loss.py
@ -0,0 +1,89 @@
+#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 os
+import pickle
+
+import paddle
+import paddle.nn as nn
+import paddle.nn.functional as F
+
+
+class CenterLoss(nn.Layer):
+    """
+    Reference: Wen et al. A Discriminative Feature Learning Approach for Deep Face Recognition. ECCV 2016.
+    """
+
+    def __init__(self,
+                 num_classes=6625,
+                 feat_dim=96,
+                 init_center=False,
+                 center_file_path=None):
+        super().__init__()
+        self.num_classes = num_classes
+        self.feat_dim = feat_dim
+        self.centers = paddle.randn(
+            shape=[self.num_classes, self.feat_dim]).astype(
+                "float64")  #random center
+
+        if init_center:
+            assert os.path.exists(
+                center_file_path
+            ), f"center path({center_file_path}) must exist when init_center is set as True."
+            with open(center_file_path, 'rb') as f:
+                char_dict = pickle.load(f)
+                for key in char_dict.keys():
+                    self.centers[key] = paddle.to_tensor(char_dict[key])
+
+    def __call__(self, predicts, batch):
+        assert isinstance(predicts, (list, tuple))
+        features, predicts = predicts
+
+        feats_reshape = paddle.reshape(
+            features, [-1, features.shape[-1]]).astype("float64")
+        label = paddle.argmax(predicts, axis=2)
+        label = paddle.reshape(label, [label.shape[0] * label.shape[1]])
+
+        batch_size = feats_reshape.shape[0]
+
+        #calc feat * feat   
+        dist1 = paddle.sum(paddle.square(feats_reshape), axis=1, keepdim=True)
+        dist1 = paddle.expand(dist1, [batch_size, self.num_classes])
+
+        #dist2 of centers
+        dist2 = paddle.sum(paddle.square(self.centers), axis=1,
+                           keepdim=True)  #num_classes
+        dist2 = paddle.expand(dist2,
+                              [self.num_classes, batch_size]).astype("float64")
+        dist2 = paddle.transpose(dist2, [1, 0])
+
+        #first x * x + y * y
+        distmat = paddle.add(dist1, dist2)
+        tmp = paddle.matmul(feats_reshape,
+                            paddle.transpose(self.centers, [1, 0]))
+        distmat = distmat - 2.0 * tmp
+
+        #generate the mask
+        classes = paddle.arange(self.num_classes).astype("int64")
+        label = paddle.expand(
+            paddle.unsqueeze(label, 1), (batch_size, self.num_classes))
+        mask = paddle.equal(
+            paddle.expand(classes, [batch_size, self.num_classes]),
+            label).astype("float64")  #get mask
+        dist = paddle.multiply(distmat, mask)
+        loss = paddle.sum(paddle.clip(dist, min=1e-12, max=1e+12)) / batch_size
+        return {'loss_center': loss}
--- a/ppocr/losses/combined_loss.py
+++ b/ppocr/losses/combined_loss.py
@ -15,6 +15,10 @@
 import paddle
 import paddle.nn as nn

+from .rec_ctc_loss import CTCLoss
+from .center_loss import CenterLoss
+from .ace_loss import ACELoss
+
 from .distillation_loss import DistillationCTCLoss
 from .distillation_loss import DistillationDMLLoss
 from .distillation_loss import DistillationDistanceLoss, DistillationDBLoss, DistillationDilaDBLoss
--- a/ppocr/losses/distillation_loss.py
+++ b/ppocr/losses/distillation_loss.py
@ -112,7 +112,7 @@ class DistillationDMLLoss(DMLLoss):
                    if isinstance(loss, dict):
                        for key in loss:
                            loss_dict["{}_{}_{}_{}_{}".format(key, pair[
-                                0], pair[1], map_name, idx)] = loss[key]
+                                0], pair[1], self.maps_name, idx)] = loss[key]
                    else:
                        loss_dict["{}_{}_{}".format(self.name, self.maps_name[
                            _c], idx)] = loss
--- a/ppocr/losses/rec_aster_loss.py
+++ b/ppocr/losses/rec_aster_loss.py
@ -0,0 +1,99 @@
+# 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 CosineEmbeddingLoss(nn.Layer):
+    def __init__(self, margin=0.):
+        super(CosineEmbeddingLoss, self).__init__()
+        self.margin = margin
+        self.epsilon = 1e-12
+
+    def forward(self, x1, x2, target):
+        similarity = paddle.fluid.layers.reduce_sum(
+            x1 * x2, dim=-1) / (paddle.norm(
+                x1, axis=-1) * paddle.norm(
+                    x2, axis=-1) + self.epsilon)
+        one_list = paddle.full_like(target, fill_value=1)
+        out = paddle.fluid.layers.reduce_mean(
+            paddle.where(
+                paddle.equal(target, one_list), 1. - similarity,
+                paddle.maximum(
+                    paddle.zeros_like(similarity), similarity - self.margin)))
+
+        return out
+
+
+class AsterLoss(nn.Layer):
+    def __init__(self,
+                 weight=None,
+                 size_average=True,
+                 ignore_index=-100,
+                 sequence_normalize=False,
+                 sample_normalize=True,
+                 **kwargs):
+        super(AsterLoss, self).__init__()
+        self.weight = weight
+        self.size_average = size_average
+        self.ignore_index = ignore_index
+        self.sequence_normalize = sequence_normalize
+        self.sample_normalize = sample_normalize
+        self.loss_sem = CosineEmbeddingLoss()
+        self.is_cosin_loss = True
+        self.loss_func_rec = nn.CrossEntropyLoss(weight=None, reduction='none')
+
+    def forward(self, predicts, batch):
+        targets = batch[1].astype("int64")
+        label_lengths = batch[2].astype('int64')
+        sem_target = batch[3].astype('float32')
+        embedding_vectors = predicts['embedding_vectors']
+        rec_pred = predicts['rec_pred']
+
+        if not self.is_cosin_loss:
+            sem_loss = paddle.sum(self.loss_sem(embedding_vectors, sem_target))
+        else:
+            label_target = paddle.ones([embedding_vectors.shape[0]])
+            sem_loss = paddle.sum(
+                self.loss_sem(embedding_vectors, sem_target, label_target))
+
+        # rec loss
+        batch_size, def_max_length = targets.shape[0], targets.shape[1]
+
+        mask = paddle.zeros([batch_size, def_max_length])
+        for i in range(batch_size):
+            mask[i, :label_lengths[i]] = 1
+        mask = paddle.cast(mask, "float32")
+        max_length = max(label_lengths)
+        assert max_length == rec_pred.shape[1]
+        targets = targets[:, :max_length]
+        mask = mask[:, :max_length]
+        rec_pred = paddle.reshape(rec_pred, [-1, rec_pred.shape[2]])
+        input = nn.functional.log_softmax(rec_pred, axis=1)
+        targets = paddle.reshape(targets, [-1, 1])
+        mask = paddle.reshape(mask, [-1, 1])
+        output = -paddle.index_sample(input, index=targets) * mask
+        output = paddle.sum(output)
+        if self.sequence_normalize:
+            output = output / paddle.sum(mask)
+        if self.sample_normalize:
+            output = output / batch_size
+
+        loss = output + sem_loss * 0.1
+        return {'loss': loss}
--- a/ppocr/losses/rec_ctc_loss.py
+++ b/ppocr/losses/rec_ctc_loss.py
@ -21,16 +21,24 @@ from paddle import nn


 class CTCLoss(nn.Layer):
-    def __init__(self, **kwargs):
+    def __init__(self, use_focal_loss=False, **kwargs):
        super(CTCLoss, self).__init__()
        self.loss_func = nn.CTCLoss(blank=0, reduction='none')
+        self.use_focal_loss = use_focal_loss

    def forward(self, predicts, batch):
+        if isinstance(predicts, (list, tuple)):
+            predicts = predicts[-1]
        predicts = predicts.transpose((1, 0, 2))
        N, B, _ = predicts.shape
        preds_lengths = paddle.to_tensor([N] * B, dtype='int64')
        labels = batch[1].astype("int32")
        label_lengths = batch[2].astype('int64')
        loss = self.loss_func(predicts, labels, preds_lengths, label_lengths)
+        if self.use_focal_loss:
+            weight = paddle.exp(-loss)
+            weight = paddle.subtract(paddle.to_tensor([1.0]), weight)
+            weight = paddle.square(weight) * self.focal_loss_alpha
+            loss = paddle.multiply(loss, weight)
        loss = loss.mean()  # sum
        return {'loss': loss}
--- a/ppocr/metrics/rec_metric.py
+++ b/ppocr/metrics/rec_metric.py
@ -13,13 +13,20 @@
 # limitations under the License.

 import Levenshtein
+import string


 class RecMetric(object):
-    def __init__(self, main_indicator='acc', **kwargs):
+    def __init__(self, main_indicator='acc', is_filter=False, **kwargs):
        self.main_indicator = main_indicator
+        self.is_filter = is_filter
        self.reset()

+    def _normalize_text(self, text):
+        text = ''.join(
+            filter(lambda x: x in (string.digits + string.ascii_letters), text))
+        return text.lower()
+
    def __call__(self, pred_label, *args, **kwargs):
        preds, labels = pred_label
        correct_num = 0
@ -28,6 +35,9 @@ class RecMetric(object):
        for (pred, pred_conf), (target, _) in zip(preds, labels):
            pred = pred.replace(" ", "")
            target = target.replace(" ", "")
+            if self.is_filter:
+                pred = self._normalize_text(pred)
+                target = self._normalize_text(target)
            norm_edit_dis += Levenshtein.distance(pred, target) / max(
                len(pred), len(target), 1)
            if pred == target:
@ -57,4 +67,3 @@ class RecMetric(object):
        self.correct_num = 0
        self.all_num = 0
        self.norm_edit_dis = 0
-        
--- a/ppocr/modeling/backbones/init.py
+++ b/ppocr/modeling/backbones/init.py
@ -28,8 +28,10 @@ def build_backbone(config, model_type):
        from .rec_mv1_enhance import MobileNetV1Enhance
        from .rec_nrtr_mtb import MTB
        from .rec_resnet_31 import ResNet31
+        from .rec_resnet_aster import ResNet_ASTER
        support_dict = [
-            'MobileNetV1Enhance', 'MobileNetV3', 'ResNet', 'ResNetFPN', 'MTB', "ResNet31"
+            'MobileNetV1Enhance', 'MobileNetV3', 'ResNet', 'ResNetFPN', 'MTB',
+            "ResNet31", "ResNet_ASTER"
        ]
    elif model_type == "e2e":
        from .e2e_resnet_vd_pg import ResNet
--- a/ppocr/modeling/backbones/rec_resnet_aster.py
+++ b/ppocr/modeling/backbones/rec_resnet_aster.py
@ -0,0 +1,140 @@
+# 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 paddle
+import paddle.nn as nn
+
+import sys
+import math
+
+
+def conv3x3(in_planes, out_planes, stride=1):
+    """3x3 convolution with padding"""
+    return nn.Conv2D(
+        in_planes,
+        out_planes,
+        kernel_size=3,
+        stride=stride,
+        padding=1,
+        bias_attr=False)
+
+
+def conv1x1(in_planes, out_planes, stride=1):
+    """1x1 convolution"""
+    return nn.Conv2D(
+        in_planes, out_planes, kernel_size=1, stride=stride, bias_attr=False)
+
+
+def get_sinusoid_encoding(n_position, feat_dim, wave_length=10000):
+    # [n_position]
+    positions = paddle.arange(0, n_position)
+    # [feat_dim]
+    dim_range = paddle.arange(0, feat_dim)
+    dim_range = paddle.pow(wave_length, 2 * (dim_range // 2) / feat_dim)
+    # [n_position, feat_dim]
+    angles = paddle.unsqueeze(
+        positions, axis=1) / paddle.unsqueeze(
+            dim_range, axis=0)
+    angles = paddle.cast(angles, "float32")
+    angles[:, 0::2] = paddle.sin(angles[:, 0::2])
+    angles[:, 1::2] = paddle.cos(angles[:, 1::2])
+    return angles
+
+
+class AsterBlock(nn.Layer):
+    def __init__(self, inplanes, planes, stride=1, downsample=None):
+        super(AsterBlock, self).__init__()
+        self.conv1 = conv1x1(inplanes, planes, stride)
+        self.bn1 = nn.BatchNorm2D(planes)
+        self.relu = nn.ReLU()
+        self.conv2 = conv3x3(planes, planes)
+        self.bn2 = nn.BatchNorm2D(planes)
+        self.downsample = downsample
+        self.stride = stride
+
+    def forward(self, x):
+        residual = x
+        out = self.conv1(x)
+        out = self.bn1(out)
+        out = self.relu(out)
+        out = self.conv2(out)
+        out = self.bn2(out)
+
+        if self.downsample is not None:
+            residual = self.downsample(x)
+        out += residual
+        out = self.relu(out)
+        return out
+
+
+class ResNet_ASTER(nn.Layer):
+    """For aster or crnn"""
+
+    def __init__(self, with_lstm=True, n_group=1, in_channels=3):
+        super(ResNet_ASTER, self).__init__()
+        self.with_lstm = with_lstm
+        self.n_group = n_group
+
+        self.layer0 = nn.Sequential(
+            nn.Conv2D(
+                in_channels,
+                32,
+                kernel_size=(3, 3),
+                stride=1,
+                padding=1,
+                bias_attr=False),
+            nn.BatchNorm2D(32),
+            nn.ReLU())
+
+        self.inplanes = 32
+        self.layer1 = self._make_layer(32, 3, [2, 2])  # [16, 50]
+        self.layer2 = self._make_layer(64, 4, [2, 2])  # [8, 25]
+        self.layer3 = self._make_layer(128, 6, [2, 1])  # [4, 25]
+        self.layer4 = self._make_layer(256, 6, [2, 1])  # [2, 25]
+        self.layer5 = self._make_layer(512, 3, [2, 1])  # [1, 25]
+
+        if with_lstm:
+            self.rnn = nn.LSTM(512, 256, direction="bidirect", num_layers=2)
+            self.out_channels = 2 * 256
+        else:
+            self.out_channels = 512
+
+    def _make_layer(self, planes, blocks, stride):
+        downsample = None
+        if stride != [1, 1] or self.inplanes != planes:
+            downsample = nn.Sequential(
+                conv1x1(self.inplanes, planes, stride), nn.BatchNorm2D(planes))
+
+        layers = []
+        layers.append(AsterBlock(self.inplanes, planes, stride, downsample))
+        self.inplanes = planes
+        for _ in range(1, blocks):
+            layers.append(AsterBlock(self.inplanes, planes))
+        return nn.Sequential(*layers)
+
+    def forward(self, x):
+        x0 = self.layer0(x)
+        x1 = self.layer1(x0)
+        x2 = self.layer2(x1)
+        x3 = self.layer3(x2)
+        x4 = self.layer4(x3)
+        x5 = self.layer5(x4)
+
+        cnn_feat = x5.squeeze(2)  # [N, c, w]
+        cnn_feat = paddle.transpose(cnn_feat, perm=[0, 2, 1])
+        if self.with_lstm:
+            rnn_feat, _ = self.rnn(cnn_feat)
+            return rnn_feat
+        else:
+            return cnn_feat
--- a/ppocr/modeling/heads/init.py
+++ b/ppocr/modeling/heads/init.py
@ -29,13 +29,14 @@ def build_head(config):
    from .rec_srn_head import SRNHead
    from .rec_nrtr_head import Transformer
    from .rec_sar_head import SARHead
+    from .rec_aster_head import AsterHead

    # cls head
    from .cls_head import ClsHead
    support_dict = [
        'DBHead', 'PSEHead', 'EASTHead', 'SASTHead', 'CTCHead', 'ClsHead',
        'AttentionHead', 'SRNHead', 'PGHead', 'Transformer',
-        'TableAttentionHead', 'SARHead'
+        'TableAttentionHead', 'SARHead', 'AsterHead'
    ]

    #table head
--- a/ppocr/modeling/heads/rec_aster_head.py
+++ b/ppocr/modeling/heads/rec_aster_head.py
@ -0,0 +1,389 @@
+# 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 sys
+
+import paddle
+from paddle import nn
+from paddle.nn import functional as F
+
+
+class AsterHead(nn.Layer):
+    def __init__(self,
+                 in_channels,
+                 out_channels,
+                 sDim,
+                 attDim,
+                 max_len_labels,
+                 time_step=25,
+                 beam_width=5,
+                 **kwargs):
+        super(AsterHead, self).__init__()
+        self.num_classes = out_channels
+        self.in_planes = in_channels
+        self.sDim = sDim
+        self.attDim = attDim
+        self.max_len_labels = max_len_labels
+        self.decoder = AttentionRecognitionHead(in_channels, out_channels, sDim,
+                                                attDim, max_len_labels)
+        self.time_step = time_step
+        self.embeder = Embedding(self.time_step, in_channels)
+        self.beam_width = beam_width
+        self.eos = self.num_classes - 1
+
+    def forward(self, x, targets=None, embed=None):
+        return_dict = {}
+        embedding_vectors = self.embeder(x)
+
+        if self.training:
+            rec_targets, rec_lengths, _ = targets
+            rec_pred = self.decoder([x, rec_targets, rec_lengths],
+                                    embedding_vectors)
+            return_dict['rec_pred'] = rec_pred
+            return_dict['embedding_vectors'] = embedding_vectors
+        else:
+            rec_pred, rec_pred_scores = self.decoder.beam_search(
+                x, self.beam_width, self.eos, embedding_vectors)
+            return_dict['rec_pred'] = rec_pred
+            return_dict['rec_pred_scores'] = rec_pred_scores
+            return_dict['embedding_vectors'] = embedding_vectors
+
+        return return_dict
+
+
+class Embedding(nn.Layer):
+    def __init__(self, in_timestep, in_planes, mid_dim=4096, embed_dim=300):
+        super(Embedding, self).__init__()
+        self.in_timestep = in_timestep
+        self.in_planes = in_planes
+        self.embed_dim = embed_dim
+        self.mid_dim = mid_dim
+        self.eEmbed = nn.Linear(
+            in_timestep * in_planes,
+            self.embed_dim)  # Embed encoder output to a word-embedding like
+
+    def forward(self, x):
+        x = paddle.reshape(x, [paddle.shape(x)[0], -1])
+        x = self.eEmbed(x)
+        return x
+
+
+class AttentionRecognitionHead(nn.Layer):
+    """
+  input: [b x 16 x 64 x in_planes]
+  output: probability sequence: [b x T x num_classes]
+  """
+
+    def __init__(self, in_channels, out_channels, sDim, attDim, max_len_labels):
+        super(AttentionRecognitionHead, self).__init__()
+        self.num_classes = out_channels  # this is the output classes. So it includes the <EOS>.
+        self.in_planes = in_channels
+        self.sDim = sDim
+        self.attDim = attDim
+        self.max_len_labels = max_len_labels
+
+        self.decoder = DecoderUnit(
+            sDim=sDim, xDim=in_channels, yDim=self.num_classes, attDim=attDim)
+
+    def forward(self, x, embed):
+        x, targets, lengths = x
+        batch_size = paddle.shape(x)[0]
+        # Decoder
+        state = self.decoder.get_initial_state(embed)
+        outputs = []
+        for i in range(max(lengths)):
+            if i == 0:
+                y_prev = paddle.full(
+                    shape=[batch_size], fill_value=self.num_classes)
+            else:
+                y_prev = targets[:, i - 1]
+            output, state = self.decoder(x, state, y_prev)
+            outputs.append(output)
+        outputs = paddle.concat([_.unsqueeze(1) for _ in outputs], 1)
+        return outputs
+
+    # inference stage.
+    def sample(self, x):
+        x, _, _ = x
+        batch_size = x.size(0)
+        # Decoder
+        state = paddle.zeros([1, batch_size, self.sDim])
+
+        predicted_ids, predicted_scores = [], []
+        for i in range(self.max_len_labels):
+            if i == 0:
+                y_prev = paddle.full(
+                    shape=[batch_size], fill_value=self.num_classes)
+            else:
+                y_prev = predicted
+
+            output, state = self.decoder(x, state, y_prev)
+            output = F.softmax(output, axis=1)
+            score, predicted = output.max(1)
+            predicted_ids.append(predicted.unsqueeze(1))
+            predicted_scores.append(score.unsqueeze(1))
+        predicted_ids = paddle.concat([predicted_ids, 1])
+        predicted_scores = paddle.concat([predicted_scores, 1])
+        # return predicted_ids.squeeze(), predicted_scores.squeeze()
+        return predicted_ids, predicted_scores
+
+    def beam_search(self, x, beam_width, eos, embed):
+        def _inflate(tensor, times, dim):
+            repeat_dims = [1] * tensor.dim()
+            repeat_dims[dim] = times
+            output = paddle.tile(tensor, repeat_dims)
+            return output
+
+        # https://github.com/IBM/pytorch-seq2seq/blob/fede87655ddce6c94b38886089e05321dc9802af/seq2seq/models/TopKDecoder.py
+        batch_size, l, d = x.shape
+        x = paddle.tile(
+            paddle.transpose(
+                x.unsqueeze(1), perm=[1, 0, 2, 3]), [beam_width, 1, 1, 1])
+        inflated_encoder_feats = paddle.reshape(
+            paddle.transpose(
+                x, perm=[1, 0, 2, 3]), [-1, l, d])
+
+        # Initialize the decoder
+        state = self.decoder.get_initial_state(embed, tile_times=beam_width)
+
+        pos_index = paddle.reshape(
+            paddle.arange(batch_size) * beam_width, shape=[-1, 1])
+
+        # Initialize the scores
+        sequence_scores = paddle.full(
+            shape=[batch_size * beam_width, 1], fill_value=-float('Inf'))
+        index = [i * beam_width for i in range(0, batch_size)]
+        sequence_scores[index] = 0.0
+
+        # Initialize the input vector
+        y_prev = paddle.full(
+            shape=[batch_size * beam_width], fill_value=self.num_classes)
+
+        # Store decisions for backtracking
+        stored_scores = list()
+        stored_predecessors = list()
+        stored_emitted_symbols = list()
+
+        for i in range(self.max_len_labels):
+            output, state = self.decoder(inflated_encoder_feats, state, y_prev)
+            state = paddle.unsqueeze(state, axis=0)
+            log_softmax_output = paddle.nn.functional.log_softmax(
+                output, axis=1)
+
+            sequence_scores = _inflate(sequence_scores, self.num_classes, 1)
+            sequence_scores += log_softmax_output
+            scores, candidates = paddle.topk(
+                paddle.reshape(sequence_scores, [batch_size, -1]),
+                beam_width,
+                axis=1)
+
+            # Reshape input = (bk, 1) and sequence_scores = (bk, 1)
+            y_prev = paddle.reshape(
+                candidates % self.num_classes, shape=[batch_size * beam_width])
+            sequence_scores = paddle.reshape(
+                scores, shape=[batch_size * beam_width, 1])
+
+            # Update fields for next timestep
+            pos_index = paddle.expand_as(pos_index, candidates)
+            predecessors = paddle.cast(
+                candidates / self.num_classes + pos_index, dtype='int64')
+            predecessors = paddle.reshape(
+                predecessors, shape=[batch_size * beam_width, 1])
+            state = paddle.index_select(
+                state, index=predecessors.squeeze(), axis=1)
+
+            # Update sequence socres and erase scores for <eos> symbol so that they aren't expanded
+            stored_scores.append(sequence_scores.clone())
+            y_prev = paddle.reshape(y_prev, shape=[-1, 1])
+            eos_prev = paddle.full_like(y_prev, fill_value=eos)
+            mask = eos_prev == y_prev
+            mask = paddle.nonzero(mask)
+            if mask.dim() > 0:
+                sequence_scores = sequence_scores.numpy()
+                mask = mask.numpy()
+                sequence_scores[mask] = -float('inf')
+                sequence_scores = paddle.to_tensor(sequence_scores)
+
+            # Cache results for backtracking
+            stored_predecessors.append(predecessors)
+            y_prev = paddle.squeeze(y_prev)
+            stored_emitted_symbols.append(y_prev)
+
+        # Do backtracking to return the optimal values
+        #====== backtrak ======#
+        # Initialize return variables given different types
+        p = list()
+        l = [[self.max_len_labels] * beam_width for _ in range(batch_size)
+             ]  # Placeholder for lengths of top-k sequences
+
+        # the last step output of the beams are not sorted
+        # thus they are sorted here
+        sorted_score, sorted_idx = paddle.topk(
+            paddle.reshape(
+                stored_scores[-1], shape=[batch_size, beam_width]),
+            beam_width)
+
+        # initialize the sequence scores with the sorted last step beam scores
+        s = sorted_score.clone()
+
+        batch_eos_found = [0] * batch_size  # the number of EOS found
+        # in the backward loop below for each batch
+        t = self.max_len_labels - 1
+        # initialize the back pointer with the sorted order of the last step beams.
+        # add pos_index for indexing variable with b*k as the first dimension.
+        t_predecessors = paddle.reshape(
+            sorted_idx + pos_index.expand_as(sorted_idx),
+            shape=[batch_size * beam_width])
+        while t >= 0:
+            # Re-order the variables with the back pointer
+            current_symbol = paddle.index_select(
+                stored_emitted_symbols[t], index=t_predecessors, axis=0)
+            t_predecessors = paddle.index_select(
+                stored_predecessors[t].squeeze(), index=t_predecessors, axis=0)
+            eos_indices = stored_emitted_symbols[t] == eos
+            eos_indices = paddle.nonzero(eos_indices)
+
+            if eos_indices.dim() > 0:
+                for i in range(eos_indices.shape[0] - 1, -1, -1):
+                    # Indices of the EOS symbol for both variables
+                    # with b*k as the first dimension, and b, k for
+                    # the first two dimensions
+                    idx = eos_indices[i]
+                    b_idx = int(idx[0] / beam_width)
+                    # The indices of the replacing position
+                    # according to the replacement strategy noted above
+                    res_k_idx = beam_width - (batch_eos_found[b_idx] %
+                                              beam_width) - 1
+                    batch_eos_found[b_idx] += 1
+                    res_idx = b_idx * beam_width + res_k_idx
+
+                    # Replace the old information in return variables
+                    # with the new ended sequence information
+                    t_predecessors[res_idx] = stored_predecessors[t][idx[0]]
+                    current_symbol[res_idx] = stored_emitted_symbols[t][idx[0]]
+                    s[b_idx, res_k_idx] = stored_scores[t][idx[0], 0]
+                    l[b_idx][res_k_idx] = t + 1
+
+            # record the back tracked results
+            p.append(current_symbol)
+            t -= 1
+
+        # Sort and re-order again as the added ended sequences may change
+        # the order (very unlikely)
+        s, re_sorted_idx = s.topk(beam_width)
+        for b_idx in range(batch_size):
+            l[b_idx] = [
+                l[b_idx][k_idx.item()] for k_idx in re_sorted_idx[b_idx, :]
+            ]
+
+        re_sorted_idx = paddle.reshape(
+            re_sorted_idx + pos_index.expand_as(re_sorted_idx),
+            [batch_size * beam_width])
+
+        # Reverse the sequences and re-order at the same time
+        # It is reversed because the backtracking happens in reverse time order
+        p = [
+            paddle.reshape(
+                paddle.index_select(step, re_sorted_idx, 0),
+                shape=[batch_size, beam_width, -1]) for step in reversed(p)
+        ]
+        p = paddle.concat(p, -1)[:, 0, :]
+        return p, paddle.ones_like(p)
+
+
+class AttentionUnit(nn.Layer):
+    def __init__(self, sDim, xDim, attDim):
+        super(AttentionUnit, self).__init__()
+
+        self.sDim = sDim
+        self.xDim = xDim
+        self.attDim = attDim
+
+        self.sEmbed = nn.Linear(sDim, attDim)
+        self.xEmbed = nn.Linear(xDim, attDim)
+        self.wEmbed = nn.Linear(attDim, 1)
+
+    def forward(self, x, sPrev):
+        batch_size, T, _ = x.shape  # [b x T x xDim]
+        x = paddle.reshape(x, [-1, self.xDim])  # [(b x T) x xDim]
+        xProj = self.xEmbed(x)  # [(b x T) x attDim]
+        xProj = paddle.reshape(xProj, [batch_size, T, -1])  # [b x T x attDim]
+
+        sPrev = sPrev.squeeze(0)
+        sProj = self.sEmbed(sPrev)  # [b x attDim]
+        sProj = paddle.unsqueeze(sProj, 1)  # [b x 1 x attDim]
+        sProj = paddle.expand(sProj,
+                              [batch_size, T, self.attDim])  # [b x T x attDim]
+
+        sumTanh = paddle.tanh(sProj + xProj)
+        sumTanh = paddle.reshape(sumTanh, [-1, self.attDim])
+
+        vProj = self.wEmbed(sumTanh)  # [(b x T) x 1]
+        vProj = paddle.reshape(vProj, [batch_size, T])
+        alpha = F.softmax(
+            vProj, axis=1)  # attention weights for each sample in the minibatch
+        return alpha
+
+
+class DecoderUnit(nn.Layer):
+    def __init__(self, sDim, xDim, yDim, attDim):
+        super(DecoderUnit, self).__init__()
+        self.sDim = sDim
+        self.xDim = xDim
+        self.yDim = yDim
+        self.attDim = attDim
+        self.emdDim = attDim
+
+        self.attention_unit = AttentionUnit(sDim, xDim, attDim)
+        self.tgt_embedding = nn.Embedding(
+            yDim + 1, self.emdDim, weight_attr=nn.initializer.Normal(
+                std=0.01))  # the last is used for <BOS>
+        self.gru = nn.GRUCell(input_size=xDim + self.emdDim, hidden_size=sDim)
+        self.fc = nn.Linear(
+            sDim,
+            yDim,
+            weight_attr=nn.initializer.Normal(std=0.01),
+            bias_attr=nn.initializer.Constant(value=0))
+        self.embed_fc = nn.Linear(300, self.sDim)
+
+    def get_initial_state(self, embed, tile_times=1):
+        assert embed.shape[1] == 300
+        state = self.embed_fc(embed)  # N * sDim
+        if tile_times != 1:
+            state = state.unsqueeze(1)
+            trans_state = paddle.transpose(state, perm=[1, 0, 2])
+            state = paddle.tile(trans_state, repeat_times=[tile_times, 1, 1])
+            trans_state = paddle.transpose(state, perm=[1, 0, 2])
+            state = paddle.reshape(trans_state, shape=[-1, self.sDim])
+        state = state.unsqueeze(0)  # 1 * N * sDim
+        return state
+
+    def forward(self, x, sPrev, yPrev):
+        # x: feature sequence from the image decoder.
+        batch_size, T, _ = x.shape
+        alpha = self.attention_unit(x, sPrev)
+        context = paddle.squeeze(paddle.matmul(alpha.unsqueeze(1), x), axis=1)
+        yPrev = paddle.cast(yPrev, dtype="int64")
+        yProj = self.tgt_embedding(yPrev)
+
+        concat_context = paddle.concat([yProj, context], 1)
+        concat_context = paddle.squeeze(concat_context, 1)
+        sPrev = paddle.squeeze(sPrev, 0)
+        output, state = self.gru(concat_context, sPrev)
+        output = paddle.squeeze(output, axis=1)
+        output = self.fc(output)
+        return output, state
--- a/ppocr/modeling/heads/rec_ctc_head.py
+++ b/ppocr/modeling/heads/rec_ctc_head.py
@ -38,6 +38,7 @@ class CTCHead(nn.Layer):
                 out_channels,
                 fc_decay=0.0004,
                 mid_channels=None,
+                 return_feats=False,
                 **kwargs):
        super(CTCHead, self).__init__()
        if mid_channels is None:
@ -66,14 +67,22 @@ class CTCHead(nn.Layer):
                bias_attr=bias_attr2)
        self.out_channels = out_channels
        self.mid_channels = mid_channels
+        self.return_feats = return_feats

    def forward(self, x, targets=None):
        if self.mid_channels is None:
            predicts = self.fc(x)
        else:
-            predicts = self.fc1(x)
-            predicts = self.fc2(predicts)
-            
+            x = self.fc1(x)
+            predicts = self.fc2(x)
+
+        if self.return_feats:
+            result = (x, predicts)
+        else:
+            result = predicts
+
        if not self.training:
            predicts = F.softmax(predicts, axis=2)
-        return predicts
+            result = predicts
+
+        return result
--- a/ppocr/modeling/transforms/init.py
+++ b/ppocr/modeling/transforms/init.py
@ -17,8 +17,9 @@ __all__ = ['build_transform']

 def build_transform(config):
    from .tps import TPS
+    from .stn import STN_ON

-    support_dict = ['TPS']
+    support_dict = ['TPS', 'STN_ON']

    module_name = config.pop('name')
    assert module_name in support_dict, Exception(
--- a/ppocr/modeling/transforms/stn.py
+++ b/ppocr/modeling/transforms/stn.py
@ -0,0 +1,132 @@
+# 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 math
+import paddle
+from paddle import nn, ParamAttr
+from paddle.nn import functional as F
+import numpy as np
+
+from .tps_spatial_transformer import TPSSpatialTransformer
+
+
+def conv3x3_block(in_channels, out_channels, stride=1):
+    n = 3 * 3 * out_channels
+    w = math.sqrt(2. / n)
+    conv_layer = nn.Conv2D(
+        in_channels,
+        out_channels,
+        kernel_size=3,
+        stride=stride,
+        padding=1,
+        weight_attr=nn.initializer.Normal(
+            mean=0.0, std=w),
+        bias_attr=nn.initializer.Constant(0))
+    block = nn.Sequential(conv_layer, nn.BatchNorm2D(out_channels), nn.ReLU())
+    return block
+
+
+class STN(nn.Layer):
+    def __init__(self, in_channels, num_ctrlpoints, activation='none'):
+        super(STN, self).__init__()
+        self.in_channels = in_channels
+        self.num_ctrlpoints = num_ctrlpoints
+        self.activation = activation
+        self.stn_convnet = nn.Sequential(
+            conv3x3_block(in_channels, 32),  #32x64
+            nn.MaxPool2D(
+                kernel_size=2, stride=2),
+            conv3x3_block(32, 64),  #16x32
+            nn.MaxPool2D(
+                kernel_size=2, stride=2),
+            conv3x3_block(64, 128),  # 8*16
+            nn.MaxPool2D(
+                kernel_size=2, stride=2),
+            conv3x3_block(128, 256),  # 4*8
+            nn.MaxPool2D(
+                kernel_size=2, stride=2),
+            conv3x3_block(256, 256),  # 2*4,
+            nn.MaxPool2D(
+                kernel_size=2, stride=2),
+            conv3x3_block(256, 256))  # 1*2
+        self.stn_fc1 = nn.Sequential(
+            nn.Linear(
+                2 * 256,
+                512,
+                weight_attr=nn.initializer.Normal(0, 0.001),
+                bias_attr=nn.initializer.Constant(0)),
+            nn.BatchNorm1D(512),
+            nn.ReLU())
+        fc2_bias = self.init_stn()
+        self.stn_fc2 = nn.Linear(
+            512,
+            num_ctrlpoints * 2,
+            weight_attr=nn.initializer.Constant(0.0),
+            bias_attr=nn.initializer.Assign(fc2_bias))
+
+    def init_stn(self):
+        margin = 0.01
+        sampling_num_per_side = int(self.num_ctrlpoints / 2)
+        ctrl_pts_x = np.linspace(margin, 1. - margin, sampling_num_per_side)
+        ctrl_pts_y_top = np.ones(sampling_num_per_side) * margin
+        ctrl_pts_y_bottom = np.ones(sampling_num_per_side) * (1 - margin)
+        ctrl_pts_top = np.stack([ctrl_pts_x, ctrl_pts_y_top], axis=1)
+        ctrl_pts_bottom = np.stack([ctrl_pts_x, ctrl_pts_y_bottom], axis=1)
+        ctrl_points = np.concatenate(
+            [ctrl_pts_top, ctrl_pts_bottom], axis=0).astype(np.float32)
+        if self.activation == 'none':
+            pass
+        elif self.activation == 'sigmoid':
+            ctrl_points = -np.log(1. / ctrl_points - 1.)
+        ctrl_points = paddle.to_tensor(ctrl_points)
+        fc2_bias = paddle.reshape(
+            ctrl_points, shape=[ctrl_points.shape[0] * ctrl_points.shape[1]])
+        return fc2_bias
+
+    def forward(self, x):
+        x = self.stn_convnet(x)
+        batch_size, _, h, w = x.shape
+        x = paddle.reshape(x, shape=(batch_size, -1))
+        img_feat = self.stn_fc1(x)
+        x = self.stn_fc2(0.1 * img_feat)
+        if self.activation == 'sigmoid':
+            x = F.sigmoid(x)
+        x = paddle.reshape(x, shape=[-1, self.num_ctrlpoints, 2])
+        return img_feat, x
+
+
+class STN_ON(nn.Layer):
+    def __init__(self, in_channels, tps_inputsize, tps_outputsize,
+                 num_control_points, tps_margins, stn_activation):
+        super(STN_ON, self).__init__()
+        self.tps = TPSSpatialTransformer(
+            output_image_size=tuple(tps_outputsize),
+            num_control_points=num_control_points,
+            margins=tuple(tps_margins))
+        self.stn_head = STN(in_channels=in_channels,
+                            num_ctrlpoints=num_control_points,
+                            activation=stn_activation)
+        self.tps_inputsize = tps_inputsize
+        self.out_channels = in_channels
+
+    def forward(self, image):
+        stn_input = paddle.nn.functional.interpolate(
+            image, self.tps_inputsize, mode="bilinear", align_corners=True)
+        stn_img_feat, ctrl_points = self.stn_head(stn_input)
+        x, _ = self.tps(image, ctrl_points)
+        return x
--- a/ppocr/modeling/transforms/tps.py
+++ b/ppocr/modeling/transforms/tps.py
@ -231,7 +231,8 @@ class GridGenerator(nn.Layer):
        """ Return inv_delta_C which is needed to calculate T """
        F = self.F
        hat_eye = paddle.eye(F, dtype='float64')  # F x F
-        hat_C = paddle.norm(C.reshape([1, F, 2]) - C.reshape([F, 1, 2]), axis=2) + hat_eye
+        hat_C = paddle.norm(
+            C.reshape([1, F, 2]) - C.reshape([F, 1, 2]), axis=2) + hat_eye
        hat_C = (hat_C**2) * paddle.log(hat_C)
        delta_C = paddle.concat(  # F+3 x F+3
            [
--- a/ppocr/modeling/transforms/tps_spatial_transformer.py
+++ b/ppocr/modeling/transforms/tps_spatial_transformer.py
@ -0,0 +1,152 @@
+# 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 math
+import paddle
+from paddle import nn, ParamAttr
+from paddle.nn import functional as F
+import numpy as np
+import itertools
+
+
+def grid_sample(input, grid, canvas=None):
+    input.stop_gradient = False
+    output = F.grid_sample(input, grid)
+    if canvas is None:
+        return output
+    else:
+        input_mask = paddle.ones(shape=input.shape)
+        output_mask = F.grid_sample(input_mask, grid)
+        padded_output = output * output_mask + canvas * (1 - output_mask)
+        return padded_output
+
+
+# phi(x1, x2) = r^2 * log(r), where r = ||x1 - x2||_2
+def compute_partial_repr(input_points, control_points):
+    N = input_points.shape[0]
+    M = control_points.shape[0]
+    pairwise_diff = paddle.reshape(
+        input_points, shape=[N, 1, 2]) - paddle.reshape(
+            control_points, shape=[1, M, 2])
+    # original implementation, very slow
+    # pairwise_dist = torch.sum(pairwise_diff ** 2, dim = 2) # square of distance
+    pairwise_diff_square = pairwise_diff * pairwise_diff
+    pairwise_dist = pairwise_diff_square[:, :, 0] + pairwise_diff_square[:, :,
+                                                                         1]
+    repr_matrix = 0.5 * pairwise_dist * paddle.log(pairwise_dist)
+    # fix numerical error for 0 * log(0), substitute all nan with 0
+    mask = repr_matrix != repr_matrix
+    repr_matrix[mask] = 0
+    return repr_matrix
+
+
+# output_ctrl_pts are specified, according to our task.
+def build_output_control_points(num_control_points, margins):
+    margin_x, margin_y = margins
+    num_ctrl_pts_per_side = num_control_points // 2
+    ctrl_pts_x = np.linspace(margin_x, 1.0 - margin_x, num_ctrl_pts_per_side)
+    ctrl_pts_y_top = np.ones(num_ctrl_pts_per_side) * margin_y
+    ctrl_pts_y_bottom = np.ones(num_ctrl_pts_per_side) * (1.0 - margin_y)
+    ctrl_pts_top = np.stack([ctrl_pts_x, ctrl_pts_y_top], axis=1)
+    ctrl_pts_bottom = np.stack([ctrl_pts_x, ctrl_pts_y_bottom], axis=1)
+    output_ctrl_pts_arr = np.concatenate(
+        [ctrl_pts_top, ctrl_pts_bottom], axis=0)
+    output_ctrl_pts = paddle.to_tensor(output_ctrl_pts_arr)
+    return output_ctrl_pts
+
+
+class TPSSpatialTransformer(nn.Layer):
+    def __init__(self,
+                 output_image_size=None,
+                 num_control_points=None,
+                 margins=None):
+        super(TPSSpatialTransformer, self).__init__()
+        self.output_image_size = output_image_size
+        self.num_control_points = num_control_points
+        self.margins = margins
+
+        self.target_height, self.target_width = output_image_size
+        target_control_points = build_output_control_points(num_control_points,
+                                                            margins)
+        N = num_control_points
+
+        # create padded kernel matrix
+        forward_kernel = paddle.zeros(shape=[N + 3, N + 3])
+        target_control_partial_repr = compute_partial_repr(
+            target_control_points, target_control_points)
+        target_control_partial_repr = paddle.cast(target_control_partial_repr,
+                                                  forward_kernel.dtype)
+        forward_kernel[:N, :N] = target_control_partial_repr
+        forward_kernel[:N, -3] = 1
+        forward_kernel[-3, :N] = 1
+        target_control_points = paddle.cast(target_control_points,
+                                            forward_kernel.dtype)
+        forward_kernel[:N, -2:] = target_control_points
+        forward_kernel[-2:, :N] = paddle.transpose(
+            target_control_points, perm=[1, 0])
+        # compute inverse matrix
+        inverse_kernel = paddle.inverse(forward_kernel)
+
+        # create target cordinate matrix
+        HW = self.target_height * self.target_width
+        target_coordinate = list(
+            itertools.product(
+                range(self.target_height), range(self.target_width)))
+        target_coordinate = paddle.to_tensor(target_coordinate)  # HW x 2
+        Y, X = paddle.split(
+            target_coordinate, target_coordinate.shape[1], axis=1)
+        Y = Y / (self.target_height - 1)
+        X = X / (self.target_width - 1)
+        target_coordinate = paddle.concat(
+            [X, Y], axis=1)  # convert from (y, x) to (x, y)
+        target_coordinate_partial_repr = compute_partial_repr(
+            target_coordinate, target_control_points)
+        target_coordinate_repr = paddle.concat(
+            [
+                target_coordinate_partial_repr, paddle.ones(shape=[HW, 1]),
+                target_coordinate
+            ],
+            axis=1)
+
+        # register precomputed matrices
+        self.inverse_kernel = inverse_kernel
+        self.padding_matrix = paddle.zeros(shape=[3, 2])
+        self.target_coordinate_repr = target_coordinate_repr
+        self.target_control_points = target_control_points
+
+    def forward(self, input, source_control_points):
+        assert source_control_points.ndimension() == 3
+        assert source_control_points.shape[1] == self.num_control_points
+        assert source_control_points.shape[2] == 2
+        batch_size = paddle.shape(source_control_points)[0]
+
+        self.padding_matrix = paddle.expand(
+            self.padding_matrix, shape=[batch_size, 3, 2])
+        Y = paddle.concat([source_control_points, self.padding_matrix], 1)
+        mapping_matrix = paddle.matmul(self.inverse_kernel, Y)
+        source_coordinate = paddle.matmul(self.target_coordinate_repr,
+                                          mapping_matrix)
+
+        grid = paddle.reshape(
+            source_coordinate,
+            shape=[-1, self.target_height, self.target_width, 2])
+        grid = paddle.clip(grid, 0,
+                           1)  # the source_control_points may be out of [0, 1].
+        # the input to grid_sample is normalized [-1, 1], but what we get is [0, 1]
+        grid = 2.0 * grid - 1.0
+        output_maps = grid_sample(input, grid, canvas=None)
+        return output_maps, source_coordinate
--- a/ppocr/optimizer/optimizer.py
+++ b/ppocr/optimizer/optimizer.py
@ -127,3 +127,34 @@ class RMSProp(object):
            grad_clip=self.grad_clip,
            parameters=parameters)
        return opt
+
+
+class Adadelta(object):
+    def __init__(self,
+                 learning_rate=0.001,
+                 epsilon=1e-08,
+                 rho=0.95,
+                 parameter_list=None,
+                 weight_decay=None,
+                 grad_clip=None,
+                 name=None,
+                 **kwargs):
+        self.learning_rate = learning_rate
+        self.epsilon = epsilon
+        self.rho = rho
+        self.parameter_list = parameter_list
+        self.learning_rate = learning_rate
+        self.weight_decay = weight_decay
+        self.grad_clip = grad_clip
+        self.name = name
+
+    def __call__(self, parameters):
+        opt = optim.Adadelta(
+            learning_rate=self.learning_rate,
+            epsilon=self.epsilon,
+            rho=self.rho,
+            weight_decay=self.weight_decay,
+            grad_clip=self.grad_clip,
+            name=self.name,
+            parameters=parameters)
+        return opt
--- a/ppocr/postprocess/init.py
+++ b/ppocr/postprocess/init.py
@ -18,17 +18,21 @@ from __future__ import print_function
 from __future__ import unicode_literals

 import copy
+import platform

 __all__ = ['build_post_process']

 from .db_postprocess import DBPostProcess, DistillationDBPostProcess
 from .east_postprocess import EASTPostProcess
 from .sast_postprocess import SASTPostProcess
-from .rec_postprocess import CTCLabelDecode, AttnLabelDecode, SRNLabelDecode, DistillationCTCLabelDecode, NRTRLabelDecode, \
-    TableLabelDecode, SARLabelDecode
+from .rec_postprocess import CTCLabelDecode, AttnLabelDecode, SRNLabelDecode, DistillationCTCLabelDecode, \
+    TableLabelDecode, NRTRLabelDecode, SARLabelDecode , SEEDLabelDecode
 from .cls_postprocess import ClsPostProcess
 from .pg_postprocess import PGPostProcess
-from .pse_postprocess import PSEPostProcess
+
+if platform.system() != "Windows":
+    # pse is not support in Windows
+    from .pse_postprocess import PSEPostProcess


 def build_post_process(config, global_config=None):
@ -36,7 +40,8 @@ def build_post_process(config, global_config=None):
        'DBPostProcess', 'PSEPostProcess', 'EASTPostProcess', 'SASTPostProcess',
        'CTCLabelDecode', 'AttnLabelDecode', 'ClsPostProcess', 'SRNLabelDecode',
        'PGPostProcess', 'DistillationCTCLabelDecode', 'TableLabelDecode',
-        'DistillationDBPostProcess', 'NRTRLabelDecode', 'SARLabelDecode'
+        'DistillationDBPostProcess', 'NRTRLabelDecode', 'SARLabelDecode',
+        'SEEDLabelDecode'
    ]

    config = copy.deepcopy(config)
--- a/ppocr/postprocess/rec_postprocess.py
+++ b/ppocr/postprocess/rec_postprocess.py
@ -111,6 +111,8 @@ class CTCLabelDecode(BaseRecLabelDecode):
                                             character_type, use_space_char)

    def __call__(self, preds, label=None, *args, **kwargs):
+        if isinstance(preds, tuple):
+            preds = preds[-1]
        if isinstance(preds, paddle.Tensor):
            preds = preds.numpy()
        preds_idx = preds.argmax(axis=2)
@ -308,6 +310,87 @@ class AttnLabelDecode(BaseRecLabelDecode):
        return idx


+class SEEDLabelDecode(BaseRecLabelDecode):
+    """ Convert between text-label and text-index """
+
+    def __init__(self,
+                 character_dict_path=None,
+                 character_type='ch',
+                 use_space_char=False,
+                 **kwargs):
+        super(SEEDLabelDecode, self).__init__(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 = dict_character + [self.end_str]
+        return dict_character
+
+    def get_ignored_tokens(self):
+        end_idx = self.get_beg_end_flag_idx("eos")
+        return [end_idx]
+
+    def get_beg_end_flag_idx(self, beg_or_end):
+        if beg_or_end == "sos":
+            idx = np.array(self.dict[self.beg_str])
+        elif beg_or_end == "eos":
+            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
+
+    def decode(self, text_index, text_prob=None, is_remove_duplicate=False):
+        """ convert text-index into text-label. """
+        result_list = []
+        [end_idx] = self.get_ignored_tokens()
+        batch_size = len(text_index)
+        for batch_idx in range(batch_size):
+            char_list = []
+            conf_list = []
+            for idx in range(len(text_index[batch_idx])):
+                if int(text_index[batch_idx][idx]) == int(end_idx):
+                    break
+                if is_remove_duplicate:
+                    # only for predict
+                    if idx > 0 and text_index[batch_idx][idx - 1] == text_index[
+                            batch_idx][idx]:
+                        continue
+                char_list.append(self.character[int(text_index[batch_idx][
+                    idx])])
+                if text_prob is not None:
+                    conf_list.append(text_prob[batch_idx][idx])
+                else:
+                    conf_list.append(1)
+            text = ''.join(char_list)
+            result_list.append((text, np.mean(conf_list)))
+        return result_list
+
+    def __call__(self, preds, label=None, *args, **kwargs):
+        """
+        text = self.decode(text)
+        if label is None:
+            return text
+        else:
+            label = self.decode(label, is_remove_duplicate=False)
+            return text, label
+        """
+        preds_idx = preds["rec_pred"]
+        if isinstance(preds_idx, paddle.Tensor):
+            preds_idx = preds_idx.numpy()
+        if "rec_pred_scores" in preds:
+            preds_idx = preds["rec_pred"]
+            preds_prob = preds["rec_pred_scores"]
+        else:
+            preds_idx = preds["rec_pred"].argmax(axis=2)
+            preds_prob = preds["rec_pred"].max(axis=2)
+        text = self.decode(preds_idx, preds_prob, is_remove_duplicate=False)
+        if label is None:
+            return text
+        label = self.decode(label, is_remove_duplicate=False)
+        return text, label
+
+
 class SRNLabelDecode(BaseRecLabelDecode):
    """ Convert between text-label and text-index """

--- a/requirements.txt
+++ b/requirements.txt
@ -11,4 +11,5 @@ opencv-contrib-python==4.4.0.46
 cython
 lxml
 premailer
-openpyxl
+openpyxl
+fasttext==0.9.1
--- a/tools/eval.py
+++ b/tools/eval.py
@ -54,8 +54,7 @@ def main():
            config['Architecture']["Head"]['out_channels'] = char_num

    model = build_model(config['Architecture'])
-    use_srn = config['Architecture']['algorithm'] == "SRN"
-    use_sar = config['Architecture']['algorithm'] == "SAR"
+    extra_input = config['Architecture']['algorithm'] in ["SRN", "SAR"]
    if "model_type" in config['Architecture'].keys():
        model_type = config['Architecture']['model_type']
    else:
@ -72,7 +71,7 @@ def main():

    # start eval
    metric = program.eval(model, valid_dataloader, post_process_class,
-                        eval_class, model_type, use_srn, use_sar)
+                          eval_class, model_type, extra_input)
    logger.info('metric eval ***************')
    for k, v in metric.items():
        logger.info('{}:{}'.format(k, v))
--- a/tools/export_model.py
+++ b/tools/export_model.py
@ -49,6 +49,12 @@ def export_single_model(model, arch_config, save_path, logger):
                ]
        ]
        model = to_static(model, input_spec=other_shape)
+    elif arch_config["algorithm"] == "SAR":
+        other_shape = [
+            paddle.static.InputSpec(
+                shape=[None, 3, 48, 160], dtype="float32"),
+        ]
+        model = to_static(model, input_spec=other_shape)
    else:
        infer_shape = [3, -1, -1]
        if arch_config["model_type"] == "rec":
--- a/tools/infer/predict_rec.py
+++ b/tools/infer/predict_rec.py
@ -68,6 +68,13 @@ class TextRecognizer(object):
                "character_dict_path": args.rec_char_dict_path,
                "use_space_char": args.use_space_char
            }
+        elif self.rec_algorithm == "SAR":
+            postprocess_params = {
+                'name': 'SARLabelDecode',
+                "character_type": args.rec_char_type,
+                "character_dict_path": args.rec_char_dict_path,
+                "use_space_char": args.use_space_char
+            }
        self.postprocess_op = build_post_process(postprocess_params)
        self.predictor, self.input_tensor, self.output_tensors, self.config = \
            utility.create_predictor(args, 'rec', logger)
@ -194,6 +201,41 @@ class TextRecognizer(object):
        return (norm_img, encoder_word_pos, gsrm_word_pos, gsrm_slf_attn_bias1,
                gsrm_slf_attn_bias2)

+    def resize_norm_img_sar(self, img, image_shape,
+                            width_downsample_ratio=0.25):
+        imgC, imgH, imgW_min, imgW_max = image_shape
+        h = img.shape[0]
+        w = img.shape[1]
+        valid_ratio = 1.0
+        # make sure new_width is an integral multiple of width_divisor.
+        width_divisor = int(1 / width_downsample_ratio)
+        # resize
+        ratio = w / float(h)
+        resize_w = math.ceil(imgH * ratio)
+        if resize_w % width_divisor != 0:
+            resize_w = round(resize_w / width_divisor) * width_divisor
+        if imgW_min is not None:
+            resize_w = max(imgW_min, resize_w)
+        if imgW_max is not None:
+            valid_ratio = min(1.0, 1.0 * resize_w / imgW_max)
+            resize_w = min(imgW_max, resize_w)
+        resized_image = cv2.resize(img, (resize_w, imgH))
+        resized_image = resized_image.astype('float32')
+        # norm 
+        if image_shape[0] == 1:
+            resized_image = resized_image / 255
+            resized_image = resized_image[np.newaxis, :]
+        else:
+            resized_image = resized_image.transpose((2, 0, 1)) / 255
+        resized_image -= 0.5
+        resized_image /= 0.5
+        resize_shape = resized_image.shape
+        padding_im = -1.0 * np.ones((imgC, imgH, imgW_max), dtype=np.float32)
+        padding_im[:, :, 0:resize_w] = resized_image
+        pad_shape = padding_im.shape
+
+        return padding_im, resize_shape, pad_shape, valid_ratio
+
    def __call__(self, img_list):
        img_num = len(img_list)
        # Calculate the aspect ratio of all text bars
@ -216,11 +258,19 @@ class TextRecognizer(object):
                wh_ratio = w * 1.0 / h
                max_wh_ratio = max(max_wh_ratio, wh_ratio)
            for ino in range(beg_img_no, end_img_no):
-                if self.rec_algorithm != "SRN":
+                if self.rec_algorithm != "SRN" and self.rec_algorithm != "SAR":
                    norm_img = self.resize_norm_img(img_list[indices[ino]],
                                                    max_wh_ratio)
                    norm_img = norm_img[np.newaxis, :]
                    norm_img_batch.append(norm_img)
+                elif self.rec_algorithm == "SAR":
+                    norm_img, _, _, valid_ratio = self.resize_norm_img_sar(
+                        img_list[indices[ino]], self.rec_image_shape)
+                    norm_img = norm_img[np.newaxis, :]
+                    valid_ratio = np.expand_dims(valid_ratio, axis=0)
+                    valid_ratios = []
+                    valid_ratios.append(valid_ratio)
+                    norm_img_batch.append(norm_img)
                else:
                    norm_img = self.process_image_srn(
                        img_list[indices[ino]], self.rec_image_shape, 8, 25)
@ -266,6 +316,25 @@ class TextRecognizer(object):
                if self.benchmark:
                    self.autolog.times.stamp()
                preds = {"predict": outputs[2]}
+            elif self.rec_algorithm == "SAR":
+                valid_ratios = np.concatenate(valid_ratios)
+                inputs = [
+                    norm_img_batch,
+                    valid_ratios,
+                ]
+                input_names = self.predictor.get_input_names()
+                for i in range(len(input_names)):
+                    input_tensor = self.predictor.get_input_handle(input_names[
+                        i])
+                    input_tensor.copy_from_cpu(inputs[i])
+                self.predictor.run()
+                outputs = []
+                for output_tensor in self.output_tensors:
+                    output = output_tensor.copy_to_cpu()
+                    outputs.append(output)
+                if self.benchmark:
+                    self.autolog.times.stamp()
+                preds = outputs[0]
            else:
                self.input_tensor.copy_from_cpu(norm_img_batch)
                self.predictor.run()
--- a/tools/program.py
+++ b/tools/program.py
@ -186,12 +186,13 @@ def train(config,
    model.train()

    use_srn = config['Architecture']['algorithm'] == "SRN"
-    use_nrtr = config['Architecture']['algorithm'] == "NRTR"
-    use_sar = config['Architecture']['algorithm'] == 'SAR'
+    extra_input = config['Architecture'][
+        'algorithm'] in ["SRN", "NRTR", "SAR", "SEED"]
    try:
        model_type = config['Architecture']['model_type']
    except:
        model_type = None
+    algorithm = config['Architecture']['algorithm']

    if 'start_epoch' in best_model_dict:
        start_epoch = best_model_dict['start_epoch']
@ -215,7 +216,7 @@ def train(config,
            images = batch[0]
            if use_srn:
                model_average = True
-            if use_srn or model_type == 'table' or use_nrtr or use_sar:
+            if model_type == 'table' or extra_input:
                preds = model(images, data=batch[1:])
            else:
                preds = model(images)
@ -279,8 +280,7 @@ def train(config,
                    post_process_class,
                    eval_class,
                    model_type,
-                    use_srn=use_srn,
-                    use_sar=use_sar)
+                    extra_input=extra_input)
                cur_metric_str = 'cur metric, {}'.format(', '.join(
                    ['{}: {}'.format(k, v) for k, v in cur_metric.items()]))
                logger.info(cur_metric_str)
@ -351,9 +351,8 @@ def eval(model,
         valid_dataloader,
         post_process_class,
         eval_class,
-         model_type,
-         use_srn=False,
-         use_sar=False):
+         model_type=None,
+         extra_input=False):
    model.eval()
    with paddle.no_grad():
        total_frame = 0.0
@ -366,7 +365,7 @@ def eval(model,
                break
            images = batch[0]
            start = time.time()
-            if use_srn or model_type == 'table' or use_sar:
+            if model_type == 'table' or extra_input:
                preds = model(images, data=batch[1:])
            else:
                preds = model(images)
@ -395,20 +394,6 @@ def preprocess(is_train=False):
    config = load_config(FLAGS.config)
    merge_config(FLAGS.opt)

-    # check if set use_gpu=True in paddlepaddle cpu version
-    use_gpu = config['Global']['use_gpu']
-    check_gpu(use_gpu)
-
-    alg = config['Architecture']['algorithm']
-    assert alg in [
-        'EAST', 'DB', 'SAST', 'Rosetta', 'CRNN', 'STARNet', 'RARE', 'SRN',
-        'CLS', 'PGNet', 'Distillation', 'NRTR', 'TableAttn', 'SAR', 'PSE'
-    ]
-
-    device = 'gpu:{}'.format(dist.ParallelEnv().dev_id) if use_gpu else 'cpu'
-    device = paddle.set_device(device)
-
-    config['Global']['distributed'] = dist.get_world_size() != 1
    if is_train:
        # save_config
        save_model_dir = config['Global']['save_model_dir']
@ -420,6 +405,27 @@ def preprocess(is_train=False):
    else:
        log_file = None
    logger = get_logger(name='root', log_file=log_file)
+
+    # check if set use_gpu=True in paddlepaddle cpu version
+    use_gpu = config['Global']['use_gpu']
+    check_gpu(use_gpu)
+
+    alg = config['Architecture']['algorithm']
+    assert alg in [
+        'EAST', 'DB', 'SAST', 'Rosetta', 'CRNN', 'STARNet', 'RARE', 'SRN',
+        'CLS', 'PGNet', 'Distillation', 'NRTR', 'TableAttn', 'SAR', 'PSE',
+        'SEED']
+    windows_not_support_list = ['PSE']
+    if platform.system() == "Windows" and alg in windows_not_support_list:
+        logger.warning('{} is not support in Windows now'.format(
+            windows_not_support_list))
+        sys.exit()
+
+    device = 'gpu:{}'.format(dist.ParallelEnv().dev_id) if use_gpu else 'cpu'
+    device = paddle.set_device(device)
+
+    config['Global']['distributed'] = dist.get_world_size() != 1
+
    if config['Global']['use_visualdl']:
        from visualdl import LogWriter
        save_model_dir = config['Global']['save_model_dir']