Merge branch 'dygraph' into dygraph

doc/doc_ch/recognition.md

@@ -11,6 +11,7 @@
     - [2.1 数据增强](#数据增强)
     - [2.2 通用模型训练](#通用模型训练)
     - [2.3 多语言模型训练](#多语言模型训练)
+    - [2.4 知识蒸馏训练](#知识蒸馏训练)
 - [3 评估](#评估)
 - [4 预测](#预测)
 - [5 转Inference模型测试](#Inference)
@@ -368,6 +369,13 @@ Eval:
     label_file_list: ["./train_data/french_val.txt"]
     ...
 ```
+
+<a name="知识蒸馏训练"></a>
+
+### 2.4 知识蒸馏训练
+
+PaddleOCR支持了基于知识蒸馏的文本识别模型训练过程，更多内容可以参考[知识蒸馏说明文档](./knowledge_distillation.md)。
+
 <a name="评估"></a>
 ## 3 评估
 

doc/doc_en/detection_en.md

@@ -9,6 +9,7 @@ This section uses the icdar2015 dataset as an example to introduce the training,
   * [2.1 Start Training](#21-start-training)
   * [2.2 Load Trained Model and Continue Training](#22-load-trained-model-and-continue-training)
   * [2.3 Training with New Backbone](#23-training-with-new-backbone)
+  * [2.4 Training with knowledge distillation](#24)
 - [3. Evaluation and Test](#3-evaluation-and-test)
   * [3.1 Evaluation](#31-evaluation)
   * [3.2 Test](#32-test)
@@ -174,6 +175,11 @@ After adding the four-part modules of the network, you only need to configure th
 
 **NOTE**: More details about replace Backbone and other mudule can be found in [doc](add_new_algorithm_en.md).
 
+
+### 2.4 Training with knowledge distillation
+
+Knowledge distillation is supported in PaddleOCR for text detection training process. For more details, please refer to [doc](./knowledge_distillation_en.md).
+
 ## 3. Evaluation and Test
 
 ### 3.1 Evaluation

doc/doc_en/models_list_en.md

@@ -94,6 +94,8 @@ For more supported languages, please refer to : [Multi-language model](./multi_l
 ## 4. Paddle-Lite Model
 |Version|Introduction|Model size|Detection model|Text Direction model|Recognition model|Paddle-Lite branch|
 |---|---|---|---|---|---|---|
+|PP-OCRv2|extra-lightweight chinese OCR optimized model|11M|[download link](https://paddleocr.bj.bcebos.com/PP-OCRv2/lite/ch_PP-OCRv2_det_infer_opt.nb)|[download link](https://paddleocr.bj.bcebos.com/PP-OCRv2/lite/ch_ppocr_mobile_v2.0_cls_infer_opt.nb)|[download link](https://paddleocr.bj.bcebos.com/PP-OCRv2/lite/ch_PP-OCRv2_rec_infer_opt.nb)|v2.10|
+|PP-OCRv2(slim)|extra-lightweight chinese OCR optimized model|4.6M|[download link](https://paddleocr.bj.bcebos.com/PP-OCRv2/lite/ch_PP-OCRv2_det_slim_opt.nb)|[download link](https://paddleocr.bj.bcebos.com/PP-OCRv2/lite/ch_ppocr_mobile_v2.0_cls_slim_opt.nb)|[download link](https://paddleocr.bj.bcebos.com/PP-OCRv2/lite/ch_PP-OCRv2_rec_slim_opt.nb)|v2.10|
 |PP-OCRv2|extra-lightweight chinese OCR optimized model|11M|[download link](https://paddleocr.bj.bcebos.com/PP-OCRv2/chinese/ch_PP-OCRv2_det_infer_opt.nb)|[download link](https://paddleocr.bj.bcebos.com/dygraph_v2.0/lite/ch_ppocr_mobile_v2.0_cls_opt.nb)|[download link](https://paddleocr.bj.bcebos.com/PP-OCRv2/chinese/ch_PP-OCRv2_rec_infer_opt.nb)|v2.9|
 |PP-OCRv2(slim)|extra-lightweight chinese OCR optimized model|4.9M|[download link](https://paddleocr.bj.bcebos.com/PP-OCRv2/chinese/ch_PP-OCRv2_det_slim_opt.nb)|[download link](https://paddleocr.bj.bcebos.com/dygraph_v2.0/lite/ch_ppocr_mobile_v2.0_cls_slim_opt.nb)|[download link](https://paddleocr.bj.bcebos.com/PP-OCRv2/chinese/ch_PP-OCRv2_rec_slim_opt.nb)|v2.9|
 |V2.0|ppocr_v2.0 extra-lightweight chinese OCR optimized model|7.8M|[download link](https://paddleocr.bj.bcebos.com/dygraph_v2.0/lite/ch_ppocr_mobile_v2.0_det_opt.nb)|[download link](https://paddleocr.bj.bcebos.com/dygraph_v2.0/lite/ch_ppocr_mobile_v2.0_cls_opt.nb)|[download link](https://paddleocr.bj.bcebos.com/dygraph_v2.0/lite/ch_ppocr_mobile_v2.0_rec_opt.nb)|v2.9|

doc/doc_en/quickstart_en.md

 
 # PaddleOCR Quick Start
 
-[PaddleOCR Quick Start](#paddleocr-quick-start)
-
 + [1. Install PaddleOCR Whl Package](#1-install-paddleocr-whl-package)
 * [2. Easy-to-Use](#2-easy-to-use)
   + [2.1 Use by Command Line](#21-use-by-command-line)

doc/doc_en/recognition_en.md

@@ -10,6 +10,7 @@
     - [2.1 Data Augmentation](#Data_Augmentation)
     - [2.2 General Training](#Training)
     - [2.3 Multi-language Training](#Multi_language)
+    - [2.4 Training with Knowledge Distillation](#kd)
 
 - [3. Evaluation](#EVALUATION)
 
@@ -361,6 +362,12 @@ Eval:
     ...
 ```
 
+<a name="kd"></a>
+
+### 2.4 Training with Knowledge Distillation
+
+Knowledge distillation is supported in PaddleOCR for text recognition training process. For more details, please refer to [doc](./knowledge_distillation_en.md).
+
 <a name="EVALUATION"></a>
 
 ## 3. Evalution

ppocr/data/imaug/__init__.py

@@ -22,7 +22,8 @@ from .make_shrink_map import MakeShrinkMap
 from .random_crop_data import EastRandomCropData, RandomCropImgMask
 from .make_pse_gt import MakePseGt
 
-from .rec_img_aug import RecAug, RecResizeImg, ClsResizeImg, SRNRecResizeImg, NRTRRecResizeImg, SARRecResizeImg
+from .rec_img_aug import RecAug, RecResizeImg, ClsResizeImg, \
+    SRNRecResizeImg, NRTRRecResizeImg, SARRecResizeImg, PRENResizeImg
 from .randaugment import RandAugment
 from .copy_paste import CopyPaste
 from .ColorJitter import ColorJitter
@@ -36,6 +37,9 @@ from .gen_table_mask import *
 
 from .vqa import *
 
+from .fce_aug import *
+from .fce_targets import FCENetTargets
+
 
 def transform(data, ops=None):
     """ transform """

ppocr/data/imaug/label_ops.py

@@ -785,6 +785,53 @@ class SARLabelEncode(BaseRecLabelEncode):
         return [self.padding_idx]
 
 
+class PRENLabelEncode(BaseRecLabelEncode):
+    def __init__(self,
+                 max_text_length,
+                 character_dict_path,
+                 use_space_char=False,
+                 **kwargs):
+        super(PRENLabelEncode, self).__init__(
+            max_text_length, character_dict_path, use_space_char)
+
+    def add_special_char(self, dict_character):
+        padding_str = '<PAD>'  # 0 
+        end_str = '<EOS>'  # 1
+        unknown_str = '<UNK>'  # 2
+
+        dict_character = [padding_str, end_str, unknown_str] + dict_character
+        self.padding_idx = 0
+        self.end_idx = 1
+        self.unknown_idx = 2
+
+        return dict_character
+
+    def encode(self, text):
+        if len(text) == 0 or len(text) >= self.max_text_len:
+            return None
+        if self.lower:
+            text = text.lower()
+        text_list = []
+        for char in text:
+            if char not in self.dict:
+                text_list.append(self.unknown_idx)
+            else:
+                text_list.append(self.dict[char])
+        text_list.append(self.end_idx)
+        if len(text_list) < self.max_text_len:
+            text_list += [self.padding_idx] * (
+                self.max_text_len - len(text_list))
+        return text_list
+
+    def __call__(self, data):
+        text = data['label']
+        encoded_text = self.encode(text)
+        if encoded_text is None:
+            return None
+        data['label'] = np.array(encoded_text)
+        return data
+
+
 class VQATokenLabelEncode(object):
     """
     Label encode for NLP VQA methods

ppocr/data/imaug/operators.py

@@ -23,14 +23,20 @@ import sys
 import six
 import cv2
 import numpy as np
+import math
 
 
 class DecodeImage(object):
     """ decode image """
 
-    def __init__(self, img_mode='RGB', channel_first=False, **kwargs):
+    def __init__(self,
+                 img_mode='RGB',
+                 channel_first=False,
+                 ignore_orientation=False,
+                 **kwargs):
         self.img_mode = img_mode
         self.channel_first = channel_first
+        self.ignore_orientation = ignore_orientation
 
     def __call__(self, data):
         img = data['image']
@@ -41,6 +47,10 @@ class DecodeImage(object):
             assert type(img) is bytes and len(
                 img) > 0, "invalid input 'img' in DecodeImage"
         img = np.frombuffer(img, dtype='uint8')
+        if self.ignore_orientation:
+            img = cv2.imdecode(img, cv2.IMREAD_IGNORE_ORIENTATION |
+                               cv2.IMREAD_COLOR)
+        else:
             img = cv2.imdecode(img, 1)
         if img is None:
             return None
@@ -156,6 +166,44 @@ class KeepKeys(object):
         return data_list
 
 
+class Pad(object):
+    def __init__(self, size=None, size_div=32, **kwargs):
+        if size is not None and not isinstance(size, (int, list, tuple)):
+            raise TypeError("Type of target_size is invalid. Now is {}".format(
+                type(size)))
+        if isinstance(size, int):
+            size = [size, size]
+        self.size = size
+        self.size_div = size_div
+
+    def __call__(self, data):
+
+        img = data['image']
+        img_h, img_w = img.shape[0], img.shape[1]
+        if self.size:
+            resize_h2, resize_w2 = self.size
+            assert (
+                img_h < resize_h2 and img_w < resize_w2
+            ), '(h, w) of target size should be greater than (img_h, img_w)'
+        else:
+            resize_h2 = max(
+                int(math.ceil(img.shape[0] / self.size_div) * self.size_div),
+                self.size_div)
+            resize_w2 = max(
+                int(math.ceil(img.shape[1] / self.size_div) * self.size_div),
+                self.size_div)
+        img = cv2.copyMakeBorder(
+            img,
+            0,
+            resize_h2 - img_h,
+            0,
+            resize_w2 - img_w,
+            cv2.BORDER_CONSTANT,
+            value=0)
+        data['image'] = img
+        return data
+
+
 class Resize(object):
     def __init__(self, size=(640, 640), **kwargs):
         self.size = size

ppocr/data/imaug/rec_img_aug.py

@@ -141,6 +141,25 @@ class SARRecResizeImg(object):
         return data
 
 
+class PRENResizeImg(object):
+    def __init__(self, image_shape, **kwargs):
+        """
+        Accroding to original paper's realization, it's a hard resize method here. 
+        So maybe you should optimize it to fit for your task better.
+        """
+        self.dst_h, self.dst_w = image_shape
+
+    def __call__(self, data):
+        img = data['image']
+        resized_img = cv2.resize(
+            img, (self.dst_w, self.dst_h), interpolation=cv2.INTER_LINEAR)
+        resized_img = resized_img.transpose((2, 0, 1)) / 255
+        resized_img -= 0.5
+        resized_img /= 0.5
+        data['image'] = resized_img.astype(np.float32)
+        return data
+
+
 def resize_norm_img_sar(img, image_shape, width_downsample_ratio=0.25):
     imgC, imgH, imgW_min, imgW_max = image_shape
     h = img.shape[0]

ppocr/data/simple_dataset.py

@@ -13,6 +13,7 @@
 # limitations under the License.
 import numpy as np
 import os
+import json
 import random
 import traceback
 from paddle.io import Dataset

ppocr/losses/__init__.py

@@ -24,6 +24,7 @@ from .det_db_loss import DBLoss
 from .det_east_loss import EASTLoss
 from .det_sast_loss import SASTLoss
 from .det_pse_loss import PSELoss
+from .det_fce_loss import FCELoss
 
 # rec loss
 from .rec_ctc_loss import CTCLoss
@@ -32,6 +33,7 @@ from .rec_srn_loss import SRNLoss
 from .rec_nrtr_loss import NRTRLoss
 from .rec_sar_loss import SARLoss
 from .rec_aster_loss import AsterLoss
+from .rec_pren_loss import PRENLoss
 
 # cls loss
 from .cls_loss import ClsLoss
@@ -55,10 +57,10 @@ from .vqa_token_layoutlm_loss import VQASerTokenLayoutLMLoss
 
 def build_loss(config):
     support_dict = [
-        'DBLoss', 'PSELoss', 'EASTLoss', 'SASTLoss', 'CTCLoss', 'ClsLoss',
-        'AttentionLoss', 'SRNLoss', 'PGLoss', 'CombinedLoss', 'NRTRLoss',
-        'TableAttentionLoss', 'SARLoss', 'AsterLoss', 'SDMGRLoss',
-        'VQASerTokenLayoutLMLoss', 'LossFromOutput'
+        'DBLoss', 'PSELoss', 'EASTLoss', 'SASTLoss', 'FCELoss', 'CTCLoss',
+        'ClsLoss', 'AttentionLoss', 'SRNLoss', 'PGLoss', 'CombinedLoss',
+        'NRTRLoss', 'TableAttentionLoss', 'SARLoss', 'AsterLoss', 'SDMGRLoss',
+        'VQASerTokenLayoutLMLoss', 'LossFromOutput', 'PRENLoss'
     ]
     config = copy.deepcopy(config)
     module_name = config.pop('name')

ppocr/losses/basic_loss.py

@@ -95,9 +95,15 @@ class DMLLoss(nn.Layer):
             self.act = None
 
         self.use_log = use_log
-
         self.jskl_loss = KLJSLoss(mode="js")
 
+    def _kldiv(self, x, target):
+        eps = 1.0e-10
+        loss = target * (paddle.log(target + eps) - x)
+        # batch mean loss
+        loss = paddle.sum(loss) / loss.shape[0]
+        return loss
+
     def forward(self, out1, out2):
         if self.act is not None:
             out1 = self.act(out1)
@@ -106,9 +112,8 @@ class DMLLoss(nn.Layer):
             # for recognition distillation, log is needed for feature map
             log_out1 = paddle.log(out1)
             log_out2 = paddle.log(out2)
-            loss = (F.kl_div(
-                log_out1, out2, reduction='batchmean') + F.kl_div(
-                    log_out2, out1, reduction='batchmean')) / 2.0
+            loss = (
+                self._kldiv(log_out1, out2) + self._kldiv(log_out2, out1)) / 2.0
         else:
             # for detection distillation log is not needed
             loss = self.jskl_loss(out1, out2)

ppocr/losses/det_fce_loss.py

+# copyright (c) 2022 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.
+"""
+This code is refer from:
+https://github.com/open-mmlab/mmocr/blob/main/mmocr/models/textdet/losses/fce_loss.py
+"""
+
+import numpy as np
+from paddle import nn
+import paddle
+import paddle.nn.functional as F
+from functools import partial
+
+
+def multi_apply(func, *args, **kwargs):
+    pfunc = partial(func, **kwargs) if kwargs else func
+    map_results = map(pfunc, *args)
+    return tuple(map(list, zip(*map_results)))
+
+
+class FCELoss(nn.Layer):
+    """The class for implementing FCENet loss
+    FCENet(CVPR2021): Fourier Contour Embedding for Arbitrary-shaped
+        Text Detection
+
+    [https://arxiv.org/abs/2104.10442]
+
+    Args:
+        fourier_degree (int) : The maximum Fourier transform degree k.
+        num_sample (int) : The sampling points number of regression
+            loss. If it is too small, fcenet tends to be overfitting.
+        ohem_ratio (float): the negative/positive ratio in OHEM.
+    """
+
+    def __init__(self, fourier_degree, num_sample, ohem_ratio=3.):
+        super().__init__()
+        self.fourier_degree = fourier_degree
+        self.num_sample = num_sample
+        self.ohem_ratio = ohem_ratio
+
+    def forward(self, preds, labels):
+        assert isinstance(preds, dict)
+        preds = preds['levels']
+
+        p3_maps, p4_maps, p5_maps = labels[1:]
+        assert p3_maps[0].shape[0] == 4 * self.fourier_degree + 5,\
+            'fourier degree not equal in FCEhead and FCEtarget'
+
+        # to tensor
+        gts = [p3_maps, p4_maps, p5_maps]
+        for idx, maps in enumerate(gts):
+            gts[idx] = paddle.to_tensor(np.stack(maps))
+
+        losses = multi_apply(self.forward_single, preds, gts)
+
+        loss_tr = paddle.to_tensor(0.).astype('float32')
+        loss_tcl = paddle.to_tensor(0.).astype('float32')
+        loss_reg_x = paddle.to_tensor(0.).astype('float32')
+        loss_reg_y = paddle.to_tensor(0.).astype('float32')
+        loss_all = paddle.to_tensor(0.).astype('float32')
+
+        for idx, loss in enumerate(losses):
+            loss_all += sum(loss)
+            if idx == 0:
+                loss_tr += sum(loss)
+            elif idx == 1:
+                loss_tcl += sum(loss)
+            elif idx == 2:
+                loss_reg_x += sum(loss)
+            else:
+                loss_reg_y += sum(loss)
+
+        results = dict(
+            loss=loss_all,
+            loss_text=loss_tr,
+            loss_center=loss_tcl,
+            loss_reg_x=loss_reg_x,
+            loss_reg_y=loss_reg_y, )
+        return results
+
+    def forward_single(self, pred, gt):
+        cls_pred = paddle.transpose(pred[0], (0, 2, 3, 1))
+        reg_pred = paddle.transpose(pred[1], (0, 2, 3, 1))
+        gt = paddle.transpose(gt, (0, 2, 3, 1))
+
+        k = 2 * self.fourier_degree + 1
+        tr_pred = paddle.reshape(cls_pred[:, :, :, :2], (-1, 2))
+        tcl_pred = paddle.reshape(cls_pred[:, :, :, 2:], (-1, 2))
+        x_pred = paddle.reshape(reg_pred[:, :, :, 0:k], (-1, k))
+        y_pred = paddle.reshape(reg_pred[:, :, :, k:2 * k], (-1, k))
+
+        tr_mask = gt[:, :, :, :1].reshape([-1])
+        tcl_mask = gt[:, :, :, 1:2].reshape([-1])
+        train_mask = gt[:, :, :, 2:3].reshape([-1])
+        x_map = paddle.reshape(gt[:, :, :, 3:3 + k], (-1, k))
+        y_map = paddle.reshape(gt[:, :, :, 3 + k:], (-1, k))
+
+        tr_train_mask = (train_mask * tr_mask).astype('bool')
+        tr_train_mask2 = paddle.concat(
+            [tr_train_mask.unsqueeze(1), tr_train_mask.unsqueeze(1)], axis=1)
+        # tr loss
+        loss_tr = self.ohem(tr_pred, tr_mask, train_mask)
+        # tcl loss
+        loss_tcl = paddle.to_tensor(0.).astype('float32')
+        tr_neg_mask = tr_train_mask.logical_not()
+        tr_neg_mask2 = paddle.concat(
+            [tr_neg_mask.unsqueeze(1), tr_neg_mask.unsqueeze(1)], axis=1)
+        if tr_train_mask.sum().item() > 0:
+            loss_tcl_pos = F.cross_entropy(
+                tcl_pred.masked_select(tr_train_mask2).reshape([-1, 2]),
+                tcl_mask.masked_select(tr_train_mask).astype('int64'))
+            loss_tcl_neg = F.cross_entropy(
+                tcl_pred.masked_select(tr_neg_mask2).reshape([-1, 2]),
+                tcl_mask.masked_select(tr_neg_mask).astype('int64'))
+            loss_tcl = loss_tcl_pos + 0.5 * loss_tcl_neg
+
+        # regression loss
+        loss_reg_x = paddle.to_tensor(0.).astype('float32')
+        loss_reg_y = paddle.to_tensor(0.).astype('float32')
+        if tr_train_mask.sum().item() > 0:
+            weight = (tr_mask.masked_select(tr_train_mask.astype('bool'))
+                      .astype('float32') + tcl_mask.masked_select(
+                          tr_train_mask.astype('bool')).astype('float32')) / 2
+            weight = weight.reshape([-1, 1])
+
+            ft_x, ft_y = self.fourier2poly(x_map, y_map)
+            ft_x_pre, ft_y_pre = self.fourier2poly(x_pred, y_pred)
+
+            dim = ft_x.shape[1]
+
+            tr_train_mask3 = paddle.concat(
+                [tr_train_mask.unsqueeze(1) for i in range(dim)], axis=1)
+
+            loss_reg_x = paddle.mean(weight * F.smooth_l1_loss(
+                ft_x_pre.masked_select(tr_train_mask3).reshape([-1, dim]),
+                ft_x.masked_select(tr_train_mask3).reshape([-1, dim]),
+                reduction='none'))
+            loss_reg_y = paddle.mean(weight * F.smooth_l1_loss(
+                ft_y_pre.masked_select(tr_train_mask3).reshape([-1, dim]),
+                ft_y.masked_select(tr_train_mask3).reshape([-1, dim]),
+                reduction='none'))
+
+        return loss_tr, loss_tcl, loss_reg_x, loss_reg_y
+
+    def ohem(self, predict, target, train_mask):
+
+        pos = (target * train_mask).astype('bool')
+        neg = ((1 - target) * train_mask).astype('bool')
+
+        pos2 = paddle.concat([pos.unsqueeze(1), pos.unsqueeze(1)], axis=1)
+        neg2 = paddle.concat([neg.unsqueeze(1), neg.unsqueeze(1)], axis=1)
+
+        n_pos = pos.astype('float32').sum()
+
+        if n_pos.item() > 0:
+            loss_pos = F.cross_entropy(
+                predict.masked_select(pos2).reshape([-1, 2]),
+                target.masked_select(pos).astype('int64'),
+                reduction='sum')
+            loss_neg = F.cross_entropy(
+                predict.masked_select(neg2).reshape([-1, 2]),
+                target.masked_select(neg).astype('int64'),
+                reduction='none')
+            n_neg = min(
+                int(neg.astype('float32').sum().item()),
+                int(self.ohem_ratio * n_pos.astype('float32')))
+        else:
+            loss_pos = paddle.to_tensor(0.)
+            loss_neg = F.cross_entropy(
+                predict.masked_select(neg2).reshape([-1, 2]),
+                target.masked_select(neg).astype('int64'),
+                reduction='none')
+            n_neg = 100
+        if len(loss_neg) > n_neg:
+            loss_neg, _ = paddle.topk(loss_neg, n_neg)
+
+        return (loss_pos + loss_neg.sum()) / (n_pos + n_neg).astype('float32')
+
+    def fourier2poly(self, real_maps, imag_maps):
+        """Transform Fourier coefficient maps to polygon maps.
+
+        Args:
+            real_maps (tensor): A map composed of the real parts of the
+                Fourier coefficients, whose shape is (-1, 2k+1)
+            imag_maps (tensor):A map composed of the imag parts of the
+                Fourier coefficients, whose shape is (-1, 2k+1)
+
+        Returns
+            x_maps (tensor): A map composed of the x value of the polygon
+                represented by n sample points (xn, yn), whose shape is (-1, n)
+            y_maps (tensor): A map composed of the y value of the polygon
+                represented by n sample points (xn, yn), whose shape is (-1, n)
+        """
+
+        k_vect = paddle.arange(
+            -self.fourier_degree, self.fourier_degree + 1,
+            dtype='float32').reshape([-1, 1])
+        i_vect = paddle.arange(
+            0, self.num_sample, dtype='float32').reshape([1, -1])
+
+        transform_matrix = 2 * np.pi / self.num_sample * paddle.matmul(k_vect,
+                                                                       i_vect)
+
+        x1 = paddle.einsum('ak, kn-> an', real_maps,
+                           paddle.cos(transform_matrix))
+        x2 = paddle.einsum('ak, kn-> an', imag_maps,
+                           paddle.sin(transform_matrix))
+        y1 = paddle.einsum('ak, kn-> an', real_maps,
+                           paddle.sin(transform_matrix))
+        y2 = paddle.einsum('ak, kn-> an', imag_maps,
+                           paddle.cos(transform_matrix))
+
+        x_maps = x1 - x2
+        y_maps = y1 + y2
+
+        return x_maps, y_maps

ppocr/losses/rec_pren_loss.py

+# copyright (c) 2022 PaddlePaddle Authors. All Rights Reserve.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#    http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+
+from __future__ import absolute_import
+from __future__ import division
+from __future__ import print_function
+
+from paddle import nn
+
+
+class PRENLoss(nn.Layer):
+    def __init__(self, **kwargs):
+        super(PRENLoss, self).__init__()
+        # note: 0 is padding idx
+        self.loss_func = nn.CrossEntropyLoss(reduction='mean', ignore_index=0)
+
+    def forward(self, predicts, batch):
+        loss = self.loss_func(predicts, batch[1].astype('int64'))
+        return {'loss': loss}

ppocr/metrics/__init__.py

@@ -21,7 +21,7 @@ import copy
 
 __all__ = ["build_metric"]
 
-from .det_metric import DetMetric
+from .det_metric import DetMetric, DetFCEMetric
 from .rec_metric import RecMetric
 from .cls_metric import ClsMetric
 from .e2e_metric import E2EMetric
@@ -34,7 +34,7 @@ from .vqa_token_re_metric import VQAReTokenMetric
 
 def build_metric(config):
     support_dict = [
-        "DetMetric", "RecMetric", "ClsMetric", "E2EMetric",
+        "DetMetric", "DetFCEMetric", "RecMetric", "ClsMetric", "E2EMetric",
         "DistillationMetric", "TableMetric", 'KIEMetric', 'VQASerTokenMetric',
         'VQAReTokenMetric'
     ]

ppocr/metrics/det_metric.py

@@ -16,7 +16,7 @@ from __future__ import absolute_import
 from __future__ import division
 from __future__ import print_function
 
-__all__ = ['DetMetric']
+__all__ = ['DetMetric', 'DetFCEMetric']
 
 from .eval_det_iou import DetectionIoUEvaluator
 
@@ -55,7 +55,6 @@ class DetMetric(object):
             result = self.evaluator.evaluate_image(gt_info_list, det_info_list)
             self.results.append(result)
 
-
     def get_metric(self):
         """
         return metrics {
@@ -71,3 +70,85 @@ class DetMetric(object):
 
     def reset(self):
         self.results = []  # clear results
+
+
+class DetFCEMetric(object):
+    def __init__(self, main_indicator='hmean', **kwargs):
+        self.evaluator = DetectionIoUEvaluator()
+        self.main_indicator = main_indicator
+        self.reset()
+
+    def __call__(self, preds, batch, **kwargs):
+        '''
+       batch: a list produced by dataloaders.
+           image: np.ndarray  of shape (N, C, H, W).
+           ratio_list: np.ndarray  of shape(N,2)
+           polygons: np.ndarray  of shape (N, K, 4, 2), the polygons of objective regions.
+           ignore_tags: np.ndarray  of shape (N, K), indicates whether a region is ignorable or not.
+       preds: a list of dict produced by post process
+            points: np.ndarray of shape (N, K, 4, 2), the polygons of objective regions.
+       '''
+        gt_polyons_batch = batch[2]
+        ignore_tags_batch = batch[3]
+
+        for pred, gt_polyons, ignore_tags in zip(preds, gt_polyons_batch,
+                                                 ignore_tags_batch):
+            # prepare gt
+            gt_info_list = [{
+                'points': gt_polyon,
+                'text': '',
+                'ignore': ignore_tag
+            } for gt_polyon, ignore_tag in zip(gt_polyons, ignore_tags)]
+            # prepare det
+            det_info_list = [{
+                'points': det_polyon,
+                'text': '',
+                'score': score
+            } for det_polyon, score in zip(pred['points'], pred['scores'])]
+
+            for score_thr in self.results.keys():
+                det_info_list_thr = [
+                    det_info for det_info in det_info_list
+                    if det_info['score'] >= score_thr
+                ]
+                result = self.evaluator.evaluate_image(gt_info_list,
+                                                       det_info_list_thr)
+                self.results[score_thr].append(result)
+
+    def get_metric(self):
+        """
+        return metrics {'heman':0,
+            'thr 0.3':'precision: 0 recall: 0 hmean: 0',
+            'thr 0.4':'precision: 0 recall: 0 hmean: 0',
+            'thr 0.5':'precision: 0 recall: 0 hmean: 0',
+            'thr 0.6':'precision: 0 recall: 0 hmean: 0',
+            'thr 0.7':'precision: 0 recall: 0 hmean: 0',
+            'thr 0.8':'precision: 0 recall: 0 hmean: 0',
+            'thr 0.9':'precision: 0 recall: 0 hmean: 0',
+            }
+        """
+        metircs = {}
+        hmean = 0
+        for score_thr in self.results.keys():
+            metirc = self.evaluator.combine_results(self.results[score_thr])
+            # for key, value in metirc.items():
+            #     metircs['{}_{}'.format(key, score_thr)] = value
+            metirc_str = 'precision:{:.5f} recall:{:.5f} hmean:{:.5f}'.format(
+                metirc['precision'], metirc['recall'], metirc['hmean'])
+            metircs['thr {}'.format(score_thr)] = metirc_str
+            hmean = max(hmean, metirc['hmean'])
+        metircs['hmean'] = hmean
+
+        self.reset()
+        return metircs
+
+    def reset(self):
+        self.results = {
+            0.3: [],
+            0.4: [],
+            0.5: [],
+            0.6: [],
+            0.7: [],
+            0.8: [],
+            0.9: []
+        }  # clear results

ppocr/modeling/backbones/__init__.py

@@ -30,9 +30,10 @@ def build_backbone(config, model_type):
         from .rec_resnet_31 import ResNet31
         from .rec_resnet_aster import ResNet_ASTER
         from .rec_micronet import MicroNet
+        from .rec_efficientb3_pren import EfficientNetb3_PREN
         support_dict = [
             'MobileNetV1Enhance', 'MobileNetV3', 'ResNet', 'ResNetFPN', 'MTB',
-            "ResNet31", "ResNet_ASTER", 'MicroNet'
+            "ResNet31", "ResNet_ASTER", 'MicroNet', 'EfficientNetb3_PREN'
         ]
     elif model_type == "e2e":
         from .e2e_resnet_vd_pg import ResNet