Merge branch 'dygraph' into dygraph

adc62fcd · topduke · GitHub · 8227ad1b · a81b88a0 · adc62fcd
Unverified Commit adc62fcd authored Aug 17, 2021 by topduke Committed by GitHub Aug 17, 2021
20 changed files
--- a/ppstructure/table/README_ch.md
+++ b/ppstructure/table/README_ch.md
+# 表格识别
+## 1. 表格识别 pipeline
+表格识别主要包含三个模型
+1. 单行文本检测-DB
+2. 单行文本识别-CRNN
+3. 表格结构和cell坐标预测-RARE
+具体流程图如下
+![tableocr_pipeline](../../doc/table/tableocr_pipeline.jpg)
+流程说明:
+1. 图片由单行文字检测模型检测到单行文字的坐标，然后送入识别模型拿到识别结果。
+2. 图片由表格结构和cell坐标预测模型拿到表格的结构信息和单元格的坐标信息。
+3. 由单行文字的坐标、识别结果和单元格的坐标一起组合出单元格的识别结果。
+4. 单元格的识别结果和表格结构一起构造表格的html字符串。
+## 2. 性能
+我们在 PubTabNet<sup>[1]</sup> 评估数据集上对算法进行了评估，性能如下
+|算法|[TEDS(Tree-Edit-Distance-based Similarity)](https://github.com/ibm-aur-nlp/PubTabNet/tree/master/src)|
+| --- | --- | 
+| EDD<sup>[2]</sup> | 88.3 | 
+| Ours | 93.32 | 
+## 3. 使用
+### 3.1 快速开始
+```python
+cd PaddleOCR/ppstructure
+# 下载模型
+mkdir inference && cd inference
+# 下载超轻量级表格英文OCR模型的检测模型并解压
+wget https://paddleocr.bj.bcebos.com/dygraph_v2.0/table/en_ppocr_mobile_v2.0_table_det_infer.tar && tar xf en_ppocr_mobile_v2.0_table_det_infer.tar
+# 下载超轻量级表格英文OCR模型的识别模型并解压
+wget https://paddleocr.bj.bcebos.com/dygraph_v2.0/table/en_ppocr_mobile_v2.0_table_rec_infer.tar && tar xf en_ppocr_mobile_v2.0_table_rec_infer.tar
+# 下载超轻量级英文表格英寸模型并解压
+wget https://paddleocr.bj.bcebos.com/dygraph_v2.0/table/en_ppocr_mobile_v2.0_table_structure_infer.tar && tar xf en_ppocr_mobile_v2.0_table_structure_infer.tar
+cd ..
+# 执行预测
+python3 table/predict_table.py --det_model_dir=inference/en_ppocr_mobile_v2.0_table_det_infer --rec_model_dir=inference/en_ppocr_mobile_v2.0_table_rec_infer --table_model_dir=inference/en_ppocr_mobile_v2.0_table_structure_infer --image_dir=../doc/table/table.jpg --rec_char_dict_path=../ppocr/utils/ppocr_keys_v1.txt --table_char_dict_path=../ppocr/utils/dict/table_structure_dict.txt --rec_char_type=ch --det_limit_side_len=736 --det_limit_type=min --output ../output/table
+```
+运行完成后，每张图片的excel表格会保存到output字段指定的目录下
+note: 上述模型是在 PubLayNet 数据集上训练的表格识别模型，仅支持英文扫描场景，如需识别其他场景需要自己训练模型后替换 `det_model_dir`,`rec_model_dir`,`table_model_dir`三个字段即可。
+### 3.2 训练
+在这一章节中，我们仅介绍表格结构模型的训练，[文字检测](../../doc/doc_ch/detection.md)和[文字识别](../../doc/doc_ch/recognition.md)的模型训练请参考对应的文档。
+#### 数据准备  
+训练数据使用公开数据集PubTabNet ([论文](https://arxiv.org/abs/1911.10683)，[下载地址](https://github.com/ibm-aur-nlp/PubTabNet))。PubTabNet数据集包含约50万张表格数据的图像，以及图像对应的html格式的注释。
+#### 启动训练  
+*如果您安装的是cpu版本，请将配置文件中的 `use_gpu` 字段修改为false*
+```shell
+# 单机单卡训练
+python3 tools/train.py -c configs/table/table_mv3.yml
+# 单机多卡训练，通过 --gpus 参数设置使用的GPU ID
+python3 -m paddle.distributed.launch --gpus '0,1,2,3' tools/train.py -c configs/table/table_mv3.yml
+```
+上述指令中，通过-c 选择训练使用configs/table/table_mv3.yml配置文件。有关配置文件的详细解释，请参考[链接](../../doc/doc_ch/config.md)。
+#### 断点训练
+如果训练程序中断，如果希望加载训练中断的模型从而恢复训练，可以通过指定Global.checkpoints指定要加载的模型路径：
+```shell
+python3 tools/train.py -c configs/table/table_mv3.yml -o Global.checkpoints=./your/trained/model
+```
+**注意**：`Global.checkpoints`的优先级高于`Global.pretrain_weights`的优先级，即同时指定两个参数时，优先加载`Global.checkpoints`指定的模型，如果`Global.checkpoints`指定的模型路径有误，会加载`Global.pretrain_weights`指定的模型。
+### 3.3 评估
+表格使用 [TEDS(Tree-Edit-Distance-based Similarity)](https://github.com/ibm-aur-nlp/PubTabNet/tree/master/src) 作为模型的评估指标。在进行模型评估之前，需要将pipeline中的三个模型分别导出为inference模型(我们已经提供好)，还需要准备评估的gt， gt示例如下:
+```json
+{"PMC4289340_004_00.png": [
+  ["<html>", "<body>", "<table>", "<thead>", "<tr>", "<td>", "</td>", "<td>", "</td>", "<td>", "</td>", "</tr>", "</thead>", "<tbody>", "<tr>", "<td>", "</td>", "<td>", "</td>", "<td>", "</td>", "</tr>",  "</tbody>", "</table>", "</body>", "</html>"], 
+  [[1, 4, 29, 13], [137, 4, 161, 13], [215, 4, 236, 13], [1, 17, 30, 27], [137, 17, 147, 27], [215, 17, 225, 27]], 
+  [["<b>", "F", "e", "a", "t", "u", "r", "e", "</b>"], ["<b>", "G", "b", "3", " ", "+", "</b>"], ["<b>", "G", "b", "3", " ", "-", "</b>"], ["<b>", "P", "a", "t", "i", "e", "n", "t", "s", "</b>"], ["6", "2"], ["4", "5"]]
+]}
+```
+json 中，key为图片名，value为对应的gt，gt是一个由三个item组成的list，每个item分别为
+1. 表格结构的html字符串list
+2. 每个cell的坐标 (不包括cell里文字为空的)
+3. 每个cell里的文字信息 (不包括cell里文字为空的)
+准备完成后使用如下命令进行评估，评估完成后会输出teds指标。
+```python
+cd PaddleOCR/ppstructure
+python3 table/eval_table.py --det_model_dir=path/to/det_model_dir --rec_model_dir=path/to/rec_model_dir --table_model_dir=path/to/table_model_dir --image_dir=../doc/table/1.png --rec_char_dict_path=../ppocr/utils/dict/table_dict.txt --table_char_dict_path=../ppocr/utils/dict/table_structure_dict.txt --rec_char_type=EN --det_limit_side_len=736 --det_limit_type=min --gt_path=path/to/gt.json
+```
+如使用PubLatNet评估数据集，将会输出
+```bash
+teds: 93.32
+```
+### 3.4 预测
+```python
+cd PaddleOCR/ppstructure
+python3 table/predict_table.py --det_model_dir=path/to/det_model_dir --rec_model_dir=path/to/rec_model_dir --table_model_dir=path/to/table_model_dir --image_dir=../doc/table/1.png --rec_char_dict_path=../ppocr/utils/dict/table_dict.txt --table_char_dict_path=../ppocr/utils/dict/table_structure_dict.txt --rec_char_type=EN --det_limit_side_len=736 --det_limit_type=min --output ../output/table
+```
+Reference
+1. https://github.com/ibm-aur-nlp/PubTabNet
+2. https://arxiv.org/pdf/1911.10683
\ No newline at end of file
--- a/ppstructure/table/__init__.py
+++ b/ppstructure/table/__init__.py
+# copyright (c) 2020 PaddlePaddle Authors. All Rights Reserve.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#    http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
--- a/ppstructure/table/eval_table.py
+++ b/ppstructure/table/eval_table.py
+# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+import os
+import sys
+__dir__ = os.path.dirname(os.path.abspath(__file__))
+sys.path.append(__dir__)
+sys.path.append(os.path.abspath(os.path.join(__dir__, '../..')))
+import cv2
+import json
+from tqdm import tqdm
+from ppstructure.table.table_metric import TEDS
+from ppstructure.table.predict_table import TableSystem
+from ppstructure.utility import init_args
+from ppocr.utils.logging import get_logger
+logger = get_logger()
+def parse_args():
+    parser = init_args()
+    parser.add_argument("--gt_path", type=str)
+    return parser.parse_args()
+def main(gt_path, img_root, args):
+    teds = TEDS(n_jobs=16)
+    text_sys = TableSystem(args)
+    jsons_gt = json.load(open(gt_path))  # gt
+    pred_htmls = []
+    gt_htmls = []
+    for img_name in tqdm(jsons_gt):
+        # read image
+        img = cv2.imread(os.path.join(img_root,img_name))
+        pred_html = text_sys(img)
+        pred_htmls.append(pred_html)
+        gt_structures, gt_bboxes, gt_contents = jsons_gt[img_name]
+        gt_html, gt = get_gt_html(gt_structures, gt_contents)
+        gt_htmls.append(gt_html)
+    scores = teds.batch_evaluate_html(gt_htmls, pred_htmls)
+    logger.info('teds:', sum(scores) / len(scores))
+def get_gt_html(gt_structures, gt_contents):
+    end_html = []
+    td_index = 0
+    for tag in gt_structures:
+        if '</td>' in tag:
+            if gt_contents[td_index] != []:
+                end_html.extend(gt_contents[td_index])
+            end_html.append(tag)
+            td_index += 1
+        else:
+            end_html.append(tag)
+    return ''.join(end_html), end_html
+if __name__ == '__main__':
+    args = parse_args()
+    main(args.gt_path,args.image_dir, args)
--- a/ppstructure/table/matcher.py
+++ b/ppstructure/table/matcher.py
+import json
+def distance(box_1, box_2):
+        x1, y1, x2, y2 = box_1
+        x3, y3, x4, y4 = box_2
+        dis = abs(x3 - x1) + abs(y3 - y1) + abs(x4- x2) + abs(y4 - y2)
+        dis_2 = abs(x3 - x1) + abs(y3 - y1)
+        dis_3 = abs(x4- x2) + abs(y4 - y2)
+        return dis + min(dis_2, dis_3)
+def compute_iou(rec1, rec2):
+    """
+    computing IoU
+    :param rec1: (y0, x0, y1, x1), which reflects
+            (top, left, bottom, right)
+    :param rec2: (y0, x0, y1, x1)
+    :return: scala value of IoU
+    """
+    # computing area of each rectangles
+    S_rec1 = (rec1[2] - rec1[0]) * (rec1[3] - rec1[1])
+    S_rec2 = (rec2[2] - rec2[0]) * (rec2[3] - rec2[1])
+    # computing the sum_area
+    sum_area = S_rec1 + S_rec2
+    # find the each edge of intersect rectangle
+    left_line = max(rec1[1], rec2[1])
+    right_line = min(rec1[3], rec2[3])
+    top_line = max(rec1[0], rec2[0])
+    bottom_line = min(rec1[2], rec2[2])
+    # judge if there is an intersect
+    if left_line >= right_line or top_line >= bottom_line:
+        return 0.0
+    else:
+        intersect = (right_line - left_line) * (bottom_line - top_line)
+        return (intersect / (sum_area - intersect))*1.0
+def matcher_merge(ocr_bboxes, pred_bboxes):
+    all_dis = []
+    ious = []
+    matched = {}
+    for i, gt_box in enumerate(ocr_bboxes):
+        distances = []
+        for j, pred_box in enumerate(pred_bboxes):
+            # compute l1 distence and IOU between two boxes
+            distances.append((distance(gt_box, pred_box), 1. - compute_iou(gt_box, pred_box)))
+        sorted_distances = distances.copy()
+        # select nearest cell
+        sorted_distances = sorted(sorted_distances, key = lambda item: (item[1], item[0])) 
+        if distances.index(sorted_distances[0]) not in matched.keys(): 
+            matched[distances.index(sorted_distances[0])] = [i]
+        else:
+            matched[distances.index(sorted_distances[0])].append(i)
+    return matched#, sum(ious) / len(ious)
+def complex_num(pred_bboxes):
+    complex_nums = []
+    for bbox in pred_bboxes:
+        distances = []
+        temp_ious = []
+        for pred_bbox in pred_bboxes:
+            if bbox != pred_bbox:
+                distances.append(distance(bbox, pred_bbox))
+                temp_ious.append(compute_iou(bbox, pred_bbox))
+        complex_nums.append(temp_ious[distances.index(min(distances))])
+    return sum(complex_nums) / len(complex_nums)
+def get_rows(pred_bboxes):
+    pre_bbox = pred_bboxes[0]
+    res = []
+    step = 0
+    for i in range(len(pred_bboxes)):
+        bbox = pred_bboxes[i]
+        if bbox[1] - pre_bbox[1] > 2 or bbox[0] - pre_bbox[0] < 0:
+            break
+        else:
+            res.append(bbox)
+            step += 1
+    for i in range(step):
+        pred_bboxes.pop(0)
+    return res, pred_bboxes
+def refine_rows(pred_bboxes): # 微调整行的框，使在一条水平线上
+    ys_1 = []
+    ys_2 = []
+    for box in pred_bboxes:
+        ys_1.append(box[1])
+        ys_2.append(box[3])
+    min_y_1 = sum(ys_1) / len(ys_1)
+    min_y_2 = sum(ys_2) / len(ys_2)
+    re_boxes = []
+    for box in pred_bboxes:
+        box[1] = min_y_1
+        box[3] = min_y_2
+        re_boxes.append(box)
+    return re_boxes
+def matcher_refine_row(gt_bboxes, pred_bboxes):
+    before_refine_pred_bboxes = pred_bboxes.copy()
+    pred_bboxes = []
+    while(len(before_refine_pred_bboxes) != 0):
+        row_bboxes, before_refine_pred_bboxes = get_rows(before_refine_pred_bboxes)
+        print(row_bboxes)
+        pred_bboxes.extend(refine_rows(row_bboxes))
+    all_dis = []
+    ious = []
+    matched = {}
+    for i, gt_box in enumerate(gt_bboxes):
+        distances = []
+        #temp_ious = []
+        for j, pred_box in enumerate(pred_bboxes):
+            distances.append(distance(gt_box, pred_box))
+            #temp_ious.append(compute_iou(gt_box, pred_box))
+        #all_dis.append(min(distances))
+        #ious.append(temp_ious[distances.index(min(distances))])
+        if distances.index(min(distances)) not in matched.keys(): 
+            matched[distances.index(min(distances))] = [i]
+        else:
+            matched[distances.index(min(distances))].append(i)
+    return matched#, sum(ious) / len(ious)
+#先挑选出一行，再进行匹配
+def matcher_structure_1(gt_bboxes, pred_bboxes_rows, pred_bboxes):
+    gt_box_index = 0
+    delete_gt_bboxes = gt_bboxes.copy()
+    match_bboxes_ready = []
+    matched = {}
+    while(len(delete_gt_bboxes) != 0):
+        row_bboxes, delete_gt_bboxes = get_rows(delete_gt_bboxes)
+        row_bboxes = sorted(row_bboxes, key = lambda key: key[0])
+        if len(pred_bboxes_rows) > 0:
+            match_bboxes_ready.extend(pred_bboxes_rows.pop(0))
+        print(row_bboxes)
+        for i, gt_box in enumerate(row_bboxes):
+            #print(gt_box)
+            pred_distances = []
+            distances = []  
+            for pred_bbox in pred_bboxes:
+                pred_distances.append(distance(gt_box, pred_bbox))
+            for j, pred_box in enumerate(match_bboxes_ready):
+                distances.append(distance(gt_box, pred_box))
+            index = pred_distances.index(min(distances))
+            #print('index', index)
+            if index not in matched.keys(): 
+                matched[index] = [gt_box_index]
+            else:
+                matched[index].append(gt_box_index)
+            gt_box_index += 1
+    return matched
+def matcher_structure(gt_bboxes, pred_bboxes_rows, pred_bboxes):
+    '''
+    gt_bboxes: 排序后
+    pred_bboxes: 
+    '''
+    pre_bbox = gt_bboxes[0]
+    matched = {}
+    match_bboxes_ready = []
+    match_bboxes_ready.extend(pred_bboxes_rows.pop(0))
+    for i, gt_box in enumerate(gt_bboxes):
+        pred_distances = []
+        for pred_bbox in pred_bboxes:
+            pred_distances.append(distance(gt_box, pred_bbox))
+        distances = []
+        gap_pre = gt_box[1] - pre_bbox[1]
+        gap_pre_1 = gt_box[0] - pre_bbox[2]
+        #print(gap_pre, len(pred_bboxes_rows))
+        if (gap_pre_1 < 0 and len(pred_bboxes_rows) > 0):
+            match_bboxes_ready.extend(pred_bboxes_rows.pop(0))
+        if len(pred_bboxes_rows) == 1:
+            match_bboxes_ready.extend(pred_bboxes_rows.pop(0))
+        if len(match_bboxes_ready) == 0 and len(pred_bboxes_rows) > 0:
+            match_bboxes_ready.extend(pred_bboxes_rows.pop(0))
+        if len(match_bboxes_ready) == 0 and len(pred_bboxes_rows) == 0:
+            break
+        #print(match_bboxes_ready)
+        for j, pred_box in enumerate(match_bboxes_ready):
+            distances.append(distance(gt_box, pred_box))
+        index = pred_distances.index(min(distances))
+        #print(gt_box, index)
+        #match_bboxes_ready.pop(distances.index(min(distances)))
+        print(gt_box, match_bboxes_ready[distances.index(min(distances))])
+        if index not in matched.keys(): 
+            matched[index] = [i]
+        else:
+            matched[index].append(i)
+        pre_bbox = gt_box
+    return matched
--- a/ppstructure/table/predict_structure.py
+++ b/ppstructure/table/predict_structure.py
+# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+import os
+import sys
+__dir__ = os.path.dirname(os.path.abspath(__file__))
+sys.path.append(__dir__)
+sys.path.append(os.path.abspath(os.path.join(__dir__, '../..')))
+os.environ["FLAGS_allocator_strategy"] = 'auto_growth'
+import cv2
+import numpy as np
+import time
+import tools.infer.utility as utility
+from ppocr.data import create_operators, transform
+from ppocr.postprocess import build_post_process
+from ppocr.utils.logging import get_logger
+from ppocr.utils.utility import get_image_file_list, check_and_read_gif
+from ppstructure.utility import parse_args
+logger = get_logger()
+class TableStructurer(object):
+    def __init__(self, args):
+        pre_process_list = [{
+            'ResizeTableImage': {
+                'max_len': args.table_max_len
+            }
+        }, {
+            'NormalizeImage': {
+                'std': [0.229, 0.224, 0.225],
+                'mean': [0.485, 0.456, 0.406],
+                'scale': '1./255.',
+                'order': 'hwc'
+            }
+        }, {
+            'PaddingTableImage': None
+        }, {
+            'ToCHWImage': None
+        }, {
+            'KeepKeys': {
+                'keep_keys': ['image']
+            }
+        }]
+        postprocess_params = {
+            'name': 'TableLabelDecode',
+            "character_type": args.table_char_type,
+            "character_dict_path": args.table_char_dict_path,
+        }
+        self.preprocess_op = create_operators(pre_process_list)
+        self.postprocess_op = build_post_process(postprocess_params)
+        self.predictor, self.input_tensor, self.output_tensors, self.config = \
+            utility.create_predictor(args, 'table', logger)
+    def __call__(self, img):
+        ori_im = img.copy()
+        data = {'image': img}
+        data = transform(data, self.preprocess_op)
+        img = data[0]
+        if img is None:
+            return None, 0
+        img = np.expand_dims(img, axis=0)
+        img = img.copy()
+        starttime = time.time()
+        self.input_tensor.copy_from_cpu(img)
+        self.predictor.run()
+        outputs = []
+        for output_tensor in self.output_tensors:
+            output = output_tensor.copy_to_cpu()
+            outputs.append(output)
+        preds = {}
+        preds['structure_probs'] = outputs[1]
+        preds['loc_preds'] = outputs[0]
+        post_result = self.postprocess_op(preds)
+        structure_str_list = post_result['structure_str_list']
+        res_loc = post_result['res_loc']
+        imgh, imgw = ori_im.shape[0:2]
+        res_loc_final = []
+        for rno in range(len(res_loc[0])):
+            x0, y0, x1, y1 = res_loc[0][rno]
+            left = max(int(imgw * x0), 0)
+            top = max(int(imgh * y0), 0)
+            right = min(int(imgw * x1), imgw - 1)
+            bottom = min(int(imgh * y1), imgh - 1)
+            res_loc_final.append([left, top, right, bottom])
+        structure_str_list = structure_str_list[0][:-1]
+        structure_str_list = ['<html>', '<body>', '<table>'] + structure_str_list + ['</table>', '</body>', '</html>']
+        elapse = time.time() - starttime
+        return (structure_str_list, res_loc_final), elapse
+def main(args):
+    image_file_list = get_image_file_list(args.image_dir)
+    table_structurer = TableStructurer(args)
+    count = 0
+    total_time = 0
+    for image_file in image_file_list:
+        img, flag = check_and_read_gif(image_file)
+        if not flag:
+            img = cv2.imread(image_file)
+        if img is None:
+            logger.info("error in loading image:{}".format(image_file))
+            continue
+        structure_res, elapse = table_structurer(img)
+        logger.info("result: {}".format(structure_res))
+        if count > 0:
+            total_time += elapse
+        count += 1
+        logger.info("Predict time of {}: {}".format(image_file, elapse))
+if __name__ == "__main__":
+    main(parse_args())
--- a/ppstructure/table/predict_table.py
+++ b/ppstructure/table/predict_table.py
+# Copyright (c) 2020 PaddlePaddle Authors. All Rights Reserved.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#     http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+import os
+import sys
+import subprocess
+__dir__ = os.path.dirname(os.path.abspath(__file__))
+sys.path.append(__dir__)
+sys.path.append(os.path.abspath(os.path.join(__dir__, '..')))
+sys.path.append(os.path.abspath(os.path.join(__dir__, '../..')))
+os.environ["FLAGS_allocator_strategy"] = 'auto_growth'
+import cv2
+import copy
+import numpy as np
+import time
+import tools.infer.predict_rec as predict_rec
+import tools.infer.predict_det as predict_det
+from ppocr.utils.utility import get_image_file_list, check_and_read_gif
+from ppocr.utils.logging import get_logger
+from ppstructure.table.matcher import distance, compute_iou
+from ppstructure.utility import parse_args
+import ppstructure.table.predict_structure as predict_strture
+logger = get_logger()
+def expand(pix, det_box, shape):
+    x0, y0, x1, y1 = det_box
+    #     print(shape)
+    h, w, c = shape
+    tmp_x0 = x0 - pix
+    tmp_x1 = x1 + pix
+    tmp_y0 = y0 - pix
+    tmp_y1 = y1 + pix
+    x0_ = tmp_x0 if tmp_x0 >= 0 else 0
+    x1_ = tmp_x1 if tmp_x1 <= w else w
+    y0_ = tmp_y0 if tmp_y0 >= 0 else 0
+    y1_ = tmp_y1 if tmp_y1 <= h else h
+    return x0_, y0_, x1_, y1_
+class TableSystem(object):
+    def __init__(self, args, text_detector=None, text_recognizer=None):
+        self.text_detector = predict_det.TextDetector(args) if text_detector is None else text_detector
+        self.text_recognizer = predict_rec.TextRecognizer(args) if text_recognizer is None else text_recognizer
+        self.table_structurer = predict_strture.TableStructurer(args)
+    def __call__(self, img):
+        ori_im = img.copy()
+        structure_res, elapse = self.table_structurer(copy.deepcopy(img))
+        dt_boxes, elapse = self.text_detector(copy.deepcopy(img))
+        dt_boxes = sorted_boxes(dt_boxes)
+        r_boxes = []
+        for box in dt_boxes:
+            x_min = box[:, 0].min() - 1
+            x_max = box[:, 0].max() + 1
+            y_min = box[:, 1].min() - 1
+            y_max = box[:, 1].max() + 1
+            box = [x_min, y_min, x_max, y_max]
+            r_boxes.append(box)
+        dt_boxes = np.array(r_boxes)
+        logger.debug("dt_boxes num : {}, elapse : {}".format(
+            len(dt_boxes), elapse))
+        if dt_boxes is None:
+            return None, None
+        img_crop_list = []
+        for i in range(len(dt_boxes)):
+            det_box = dt_boxes[i]
+            x0, y0, x1, y1 = expand(2, det_box, ori_im.shape)
+            text_rect = ori_im[int(y0):int(y1), int(x0):int(x1), :]
+            img_crop_list.append(text_rect)
+        rec_res, elapse = self.text_recognizer(img_crop_list)
+        logger.debug("rec_res num  : {}, elapse : {}".format(
+            len(rec_res), elapse))
+        pred_html, pred = self.rebuild_table(structure_res, dt_boxes, rec_res)
+        return pred_html
+    def rebuild_table(self, structure_res, dt_boxes, rec_res):
+        pred_structures, pred_bboxes = structure_res
+        matched_index = self.match_result(dt_boxes, pred_bboxes)
+        pred_html, pred = self.get_pred_html(pred_structures, matched_index, rec_res)
+        return pred_html, pred
+    def match_result(self, dt_boxes, pred_bboxes):
+        matched = {}
+        for i, gt_box in enumerate(dt_boxes):
+            # gt_box = [np.min(gt_box[:, 0]), np.min(gt_box[:, 1]), np.max(gt_box[:, 0]), np.max(gt_box[:, 1])]
+            distances = []
+            for j, pred_box in enumerate(pred_bboxes):
+                distances.append(
+                    (distance(gt_box, pred_box), 1. - compute_iou(gt_box, pred_box)))  # 获取两两cell之间的L1距离和 1- IOU
+            sorted_distances = distances.copy()
+            # 根据距离和IOU挑选最"近"的cell
+            sorted_distances = sorted(sorted_distances, key=lambda item: (item[1], item[0]))
+            if distances.index(sorted_distances[0]) not in matched.keys():
+                matched[distances.index(sorted_distances[0])] = [i]
+            else:
+                matched[distances.index(sorted_distances[0])].append(i)
+        return matched
+    def get_pred_html(self, pred_structures, matched_index, ocr_contents):
+        end_html = []
+        td_index = 0
+        for tag in pred_structures:
+            if '</td>' in tag:
+                if td_index in matched_index.keys():
+                    b_with = False
+                    if '<b>' in ocr_contents[matched_index[td_index][0]] and len(matched_index[td_index]) > 1:
+                        b_with = True
+                        end_html.extend('<b>')
+                    for i, td_index_index in enumerate(matched_index[td_index]):
+                        content = ocr_contents[td_index_index][0]
+                        if len(matched_index[td_index]) > 1:
+                            if len(content) == 0:
+                                continue
+                            if content[0] == ' ':
+                                content = content[1:]
+                            if '<b>' in content:
+                                content = content[3:]
+                            if '</b>' in content:
+                                content = content[:-4]
+                            if len(content) == 0:
+                                continue
+                            if i != len(matched_index[td_index]) - 1 and ' ' != content[-1]:
+                                content += ' '
+                        end_html.extend(content)
+                    if b_with:
+                        end_html.extend('</b>')
+                end_html.append(tag)
+                td_index += 1
+            else:
+                end_html.append(tag)
+        return ''.join(end_html), end_html
+def sorted_boxes(dt_boxes):
+    """
+    Sort text boxes in order from top to bottom, left to right
+    args:
+        dt_boxes(array):detected text boxes with shape [4, 2]
+    return:
+        sorted boxes(array) with shape [4, 2]
+    """
+    num_boxes = dt_boxes.shape[0]
+    sorted_boxes = sorted(dt_boxes, key=lambda x: (x[0][1], x[0][0]))
+    _boxes = list(sorted_boxes)
+    for i in range(num_boxes - 1):
+        if abs(_boxes[i + 1][0][1] - _boxes[i][0][1]) < 10 and \
+                (_boxes[i + 1][0][0] < _boxes[i][0][0]):
+            tmp = _boxes[i]
+            _boxes[i] = _boxes[i + 1]
+            _boxes[i + 1] = tmp
+    return _boxes
+def to_excel(html_table, excel_path):
+    from tablepyxl import tablepyxl
+    tablepyxl.document_to_xl(html_table, excel_path)
+def main(args):
+    image_file_list = get_image_file_list(args.image_dir)
+    image_file_list = image_file_list[args.process_id::args.total_process_num]
+    os.makedirs(args.output, exist_ok=True)
+    text_sys = TableSystem(args)
+    img_num = len(image_file_list)
+    for i, image_file in enumerate(image_file_list):
+        logger.info("[{}/{}] {}".format(i, img_num, image_file))
+        img, flag = check_and_read_gif(image_file)
+        excel_path = os.path.join(args.output, os.path.basename(image_file).split('.')[0] + '.xlsx')
+        if not flag:
+            img = cv2.imread(image_file)
+        if img is None:
+            logger.error("error in loading image:{}".format(image_file))
+            continue
+        starttime = time.time()
+        pred_html = text_sys(img)
+        to_excel(pred_html, excel_path)
+        logger.info('excel saved to {}'.format(excel_path))
+        logger.info(pred_html)
+        elapse = time.time() - starttime
+        logger.info("Predict time : {:.3f}s".format(elapse))
+if __name__ == "__main__":
+    args = parse_args()
+    if args.use_mp:
+        p_list = []
+        total_process_num = args.total_process_num
+        for process_id in range(total_process_num):
+            cmd = [sys.executable, "-u"] + sys.argv + [
+                "--process_id={}".format(process_id),
+                "--use_mp={}".format(False)
+            ]
+            p = subprocess.Popen(cmd, stdout=sys.stdout, stderr=sys.stdout)
+            p_list.append(p)
+        for p in p_list:
+            p.wait()
+    else:
+        main(args)
--- a/ppstructure/table/table_metric/__init__.py
+++ b/ppstructure/table/table_metric/__init__.py
+# copyright (c) 2020 PaddlePaddle Authors. All Rights Reserve.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#    http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+__all__ = ['TEDS']
+from .table_metric import TEDS
\ No newline at end of file
--- a/ppstructure/table/table_metric/parallel.py
+++ b/ppstructure/table/table_metric/parallel.py
+from tqdm import tqdm
+from concurrent.futures import ProcessPoolExecutor, as_completed
+def parallel_process(array, function, n_jobs=16, use_kwargs=False, front_num=0):
+    """
+        A parallel version of the map function with a progress bar.
+        Args:
+            array (array-like): An array to iterate over.
+            function (function): A python function to apply to the elements of array
+            n_jobs (int, default=16): The number of cores to use
+            use_kwargs (boolean, default=False): Whether to consider the elements of array as dictionaries of
+                keyword arguments to function
+            front_num (int, default=3): The number of iterations to run serially before kicking off the parallel job.
+                Useful for catching bugs
+        Returns:
+            [function(array[0]), function(array[1]), ...]
+    """
+    # We run the first few iterations serially to catch bugs
+    if front_num > 0:
+        front = [function(**a) if use_kwargs else function(a)
+                 for a in array[:front_num]]
+    else:
+        front = []
+    # If we set n_jobs to 1, just run a list comprehension. This is useful for benchmarking and debugging.
+    if n_jobs == 1:
+        return front + [function(**a) if use_kwargs else function(a) for a in tqdm(array[front_num:])]
+    # Assemble the workers
+    with ProcessPoolExecutor(max_workers=n_jobs) as pool:
+        # Pass the elements of array into function
+        if use_kwargs:
+            futures = [pool.submit(function, **a) for a in array[front_num:]]
+        else:
+            futures = [pool.submit(function, a) for a in array[front_num:]]
+        kwargs = {
+            'total': len(futures),
+            'unit': 'it',
+            'unit_scale': True,
+            'leave': True
+        }
+        # Print out the progress as tasks complete
+        for f in tqdm(as_completed(futures), **kwargs):
+            pass
+    out = []
+    # Get the results from the futures.
+    for i, future in tqdm(enumerate(futures)):
+        try:
+            out.append(future.result())
+        except Exception as e:
+            out.append(e)
+    return front + out
--- a/ppstructure/table/table_metric/table_metric.py
+++ b/ppstructure/table/table_metric/table_metric.py
+# Copyright 2020 IBM
+# Author: peter.zhong@au1.ibm.com
+#
+# This is free software; you can redistribute it and/or modify
+# it under the terms of the Apache 2.0 License.
+#
+# This software is distributed in the hope that it will be useful,
+# but WITHOUT ANY WARRANTY; without even the implied warranty of
+# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+# Apache 2.0 License for more details.
+import distance
+from apted import APTED, Config
+from apted.helpers import Tree
+from lxml import etree, html
+from collections import deque
+from .parallel import parallel_process
+from tqdm import tqdm
+class TableTree(Tree):
+    def __init__(self, tag, colspan=None, rowspan=None, content=None, *children):
+        self.tag = tag
+        self.colspan = colspan
+        self.rowspan = rowspan
+        self.content = content
+        self.children = list(children)
+    def bracket(self):
+        """Show tree using brackets notation"""
+        if self.tag == 'td':
+            result = '"tag": %s, "colspan": %d, "rowspan": %d, "text": %s' % \
+                     (self.tag, self.colspan, self.rowspan, self.content)
+        else:
+            result = '"tag": %s' % self.tag
+        for child in self.children:
+            result += child.bracket()
+        return "{{{}}}".format(result)
+class CustomConfig(Config):
+    @staticmethod
+    def maximum(*sequences):
+        """Get maximum possible value
+        """
+        return max(map(len, sequences))
+    def normalized_distance(self, *sequences):
+        """Get distance from 0 to 1
+        """
+        return float(distance.levenshtein(*sequences)) / self.maximum(*sequences)
+    def rename(self, node1, node2):
+        """Compares attributes of trees"""
+        #print(node1.tag)
+        if (node1.tag != node2.tag) or (node1.colspan != node2.colspan) or (node1.rowspan != node2.rowspan):
+            return 1.
+        if node1.tag == 'td':
+            if node1.content or node2.content:
+                #print(node1.content, )
+                return self.normalized_distance(node1.content, node2.content)
+        return 0.
+class CustomConfig_del_short(Config):
+    @staticmethod
+    def maximum(*sequences):
+        """Get maximum possible value
+        """
+        return max(map(len, sequences))
+    def normalized_distance(self, *sequences):
+        """Get distance from 0 to 1
+        """
+        return float(distance.levenshtein(*sequences)) / self.maximum(*sequences)
+    def rename(self, node1, node2):
+        """Compares attributes of trees"""
+        if (node1.tag != node2.tag) or (node1.colspan != node2.colspan) or (node1.rowspan != node2.rowspan):
+            return 1.
+        if node1.tag == 'td':
+            if node1.content or node2.content:
+                #print('before')
+                #print(node1.content, node2.content)
+                #print('after')
+                node1_content = node1.content
+                node2_content = node2.content
+                if len(node1_content) < 3:
+                    node1_content = ['####']
+                if len(node2_content) < 3:
+                    node2_content = ['####']   
+                return self.normalized_distance(node1_content, node2_content)
+        return 0.
+class CustomConfig_del_block(Config):
+    @staticmethod
+    def maximum(*sequences):
+        """Get maximum possible value
+        """
+        return max(map(len, sequences))
+    def normalized_distance(self, *sequences):
+        """Get distance from 0 to 1
+        """
+        return float(distance.levenshtein(*sequences)) / self.maximum(*sequences)
+    def rename(self, node1, node2):
+        """Compares attributes of trees"""
+        if (node1.tag != node2.tag) or (node1.colspan != node2.colspan) or (node1.rowspan != node2.rowspan):
+            return 1.
+        if node1.tag == 'td':
+            if node1.content or node2.content:
+                node1_content = node1.content
+                node2_content = node2.content
+                while ' '  in node1_content:
+                    print(node1_content.index(' '))
+                    node1_content.pop(node1_content.index(' '))
+                while ' ' in node2_content:
+                    print(node2_content.index(' '))
+                    node2_content.pop(node2_content.index(' '))
+                return self.normalized_distance(node1_content, node2_content)
+        return 0.
+class TEDS(object):
+    ''' Tree Edit Distance basead Similarity
+    '''
+    def __init__(self, structure_only=False, n_jobs=1, ignore_nodes=None):
+        assert isinstance(n_jobs, int) and (
+            n_jobs >= 1), 'n_jobs must be an integer greather than 1'
+        self.structure_only = structure_only
+        self.n_jobs = n_jobs
+        self.ignore_nodes = ignore_nodes
+        self.__tokens__ = []
+    def tokenize(self, node):
+        ''' Tokenizes table cells
+        '''
+        self.__tokens__.append('<%s>' % node.tag)
+        if node.text is not None:
+            self.__tokens__ += list(node.text)
+        for n in node.getchildren():
+            self.tokenize(n)
+        if node.tag != 'unk':
+            self.__tokens__.append('</%s>' % node.tag)
+        if node.tag != 'td' and node.tail is not None:
+            self.__tokens__ += list(node.tail)
+    def load_html_tree(self, node, parent=None):
+        ''' Converts HTML tree to the format required by apted
+        '''
+        global __tokens__
+        if node.tag == 'td':
+            if self.structure_only:
+                cell = []
+            else:
+                self.__tokens__ = []
+                self.tokenize(node)
+                cell = self.__tokens__[1:-1].copy()
+            new_node = TableTree(node.tag,
+                                 int(node.attrib.get('colspan', '1')),
+                                 int(node.attrib.get('rowspan', '1')),
+                                 cell, *deque())
+        else:
+            new_node = TableTree(node.tag, None, None, None, *deque())
+        if parent is not None:
+            parent.children.append(new_node)
+        if node.tag != 'td':
+            for n in node.getchildren():
+                self.load_html_tree(n, new_node)
+        if parent is None:
+            return new_node
+    def evaluate(self, pred, true):
+        ''' Computes TEDS score between the prediction and the ground truth of a
+            given sample
+        '''
+        if (not pred) or (not true):
+            return 0.0
+        parser = html.HTMLParser(remove_comments=True, encoding='utf-8')
+        pred = html.fromstring(pred, parser=parser)
+        true = html.fromstring(true, parser=parser)
+        if pred.xpath('body/table') and true.xpath('body/table'):
+            pred = pred.xpath('body/table')[0]
+            true = true.xpath('body/table')[0]
+            if self.ignore_nodes:
+                etree.strip_tags(pred, *self.ignore_nodes)
+                etree.strip_tags(true, *self.ignore_nodes)
+            n_nodes_pred = len(pred.xpath(".//*"))
+            n_nodes_true = len(true.xpath(".//*"))
+            n_nodes = max(n_nodes_pred, n_nodes_true)
+            tree_pred = self.load_html_tree(pred)
+            tree_true = self.load_html_tree(true)
+            distance = APTED(tree_pred, tree_true,
+                             CustomConfig()).compute_edit_distance()
+            return 1.0 - (float(distance) / n_nodes)
+        else:
+            return 0.0
+    def batch_evaluate(self, pred_json, true_json):
+        ''' Computes TEDS score between the prediction and the ground truth of
+            a batch of samples
+            @params pred_json: {'FILENAME': 'HTML CODE', ...}
+            @params true_json: {'FILENAME': {'html': 'HTML CODE'}, ...}
+            @output: {'FILENAME': 'TEDS SCORE', ...}
+        '''
+        samples = true_json.keys()
+        if self.n_jobs == 1:
+            scores = [self.evaluate(pred_json.get(
+                filename, ''), true_json[filename]['html']) for filename in tqdm(samples)]
+        else:
+            inputs = [{'pred': pred_json.get(
+                filename, ''), 'true': true_json[filename]['html']} for filename in samples]
+            scores = parallel_process(
+                inputs, self.evaluate, use_kwargs=True, n_jobs=self.n_jobs, front_num=1)
+        scores = dict(zip(samples, scores))
+        return scores
+    def batch_evaluate_html(self, pred_htmls, true_htmls):
+        ''' Computes TEDS score between the prediction and the ground truth of
+            a batch of samples
+        '''
+        if self.n_jobs == 1:
+            scores = [self.evaluate(pred_html, true_html) for (
+                pred_html, true_html) in zip(pred_htmls, true_htmls)]
+        else:
+            inputs = [{"pred": pred_html, "true": true_html} for(
+                pred_html, true_html) in zip(pred_htmls, true_htmls)]
+            scores = parallel_process(
+                inputs, self.evaluate, use_kwargs=True, n_jobs=self.n_jobs, front_num=1)
+        return scores
+if __name__ == '__main__':
+    import json
+    import pprint
+    with open('sample_pred.json') as fp:
+        pred_json = json.load(fp)
+    with open('sample_gt.json') as fp:
+        true_json = json.load(fp)
+    teds = TEDS(n_jobs=4)
+    scores = teds.batch_evaluate(pred_json, true_json)
+    pp = pprint.PrettyPrinter()
+    pp.pprint(scores)
--- a/ppstructure/table/tablepyxl/__init__.py
+++ b/ppstructure/table/tablepyxl/__init__.py
+# copyright (c) 2020 PaddlePaddle Authors. All Rights Reserve.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#    http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
\ No newline at end of file
--- a/ppstructure/table/tablepyxl/style.py
+++ b/ppstructure/table/tablepyxl/style.py
+# This is where we handle translating css styles into openpyxl styles
+# and cascading those from parent to child in the dom.
+from openpyxl.cell import cell
+from openpyxl.styles import Font, Alignment, PatternFill, NamedStyle, Border, Side, Color
+from openpyxl.styles.fills import FILL_SOLID
+from openpyxl.styles.numbers import FORMAT_CURRENCY_USD_SIMPLE, FORMAT_PERCENTAGE
+from openpyxl.styles.colors import BLACK
+FORMAT_DATE_MMDDYYYY = 'mm/dd/yyyy'
+def colormap(color):
+    """
+    Convenience for looking up known colors
+    """
+    cmap = {'black': BLACK}
+    return cmap.get(color, color)
+def style_string_to_dict(style):
+    """
+    Convert css style string to a python dictionary
+    """
+    def clean_split(string, delim):
+        return (s.strip() for s in string.split(delim))
+    styles = [clean_split(s, ":") for s in style.split(";") if ":" in s]
+    return dict(styles)
+def get_side(style, name):
+    return {'border_style': style.get('border-{}-style'.format(name)),
+            'color': colormap(style.get('border-{}-color'.format(name)))}
+known_styles = {}
+def style_dict_to_named_style(style_dict, number_format=None):
+    """
+    Change css style (stored in a python dictionary) to openpyxl NamedStyle
+    """
+    style_and_format_string = str({
+        'style_dict': style_dict,
+        'parent': style_dict.parent,
+        'number_format': number_format,
+    })
+    if style_and_format_string not in known_styles:
+        # Font
+        font = Font(bold=style_dict.get('font-weight') == 'bold',
+                    color=style_dict.get_color('color', None),
+                    size=style_dict.get('font-size'))
+        # Alignment
+        alignment = Alignment(horizontal=style_dict.get('text-align', 'general'),
+                              vertical=style_dict.get('vertical-align'),
+                              wrap_text=style_dict.get('white-space', 'nowrap') == 'normal')
+        # Fill
+        bg_color = style_dict.get_color('background-color')
+        fg_color = style_dict.get_color('foreground-color', Color())
+        fill_type = style_dict.get('fill-type')
+        if bg_color and bg_color != 'transparent':
+            fill = PatternFill(fill_type=fill_type or FILL_SOLID,
+                               start_color=bg_color,
+                               end_color=fg_color)
+        else:
+            fill = PatternFill()
+        # Border
+        border = Border(left=Side(**get_side(style_dict, 'left')),
+                        right=Side(**get_side(style_dict, 'right')),
+                        top=Side(**get_side(style_dict, 'top')),
+                        bottom=Side(**get_side(style_dict, 'bottom')),
+                        diagonal=Side(**get_side(style_dict, 'diagonal')),
+                        diagonal_direction=None,
+                        outline=Side(**get_side(style_dict, 'outline')),
+                        vertical=None,
+                        horizontal=None)
+        name = 'Style {}'.format(len(known_styles) + 1)
+        pyxl_style = NamedStyle(name=name, font=font, fill=fill, alignment=alignment, border=border,
+                                number_format=number_format)
+        known_styles[style_and_format_string] = pyxl_style
+    return known_styles[style_and_format_string]
+class StyleDict(dict):
+    """
+    It's like a dictionary, but it looks for items in the parent dictionary
+    """
+    def __init__(self, *args, **kwargs):
+        self.parent = kwargs.pop('parent', None)
+        super(StyleDict, self).__init__(*args, **kwargs)
+    def __getitem__(self, item):
+        if item in self:
+            return super(StyleDict, self).__getitem__(item)
+        elif self.parent:
+            return self.parent[item]
+        else:
+            raise KeyError('{} not found'.format(item))
+    def __hash__(self):
+        return hash(tuple([(k, self.get(k)) for k in self._keys()]))
+    # Yielding the keys avoids creating unnecessary data structures
+    # and happily works with both python2 and python3 where the
+    # .keys() method is a dictionary_view in python3 and a list in python2.
+    def _keys(self):
+        yielded = set()
+        for k in self.keys():
+            yielded.add(k)
+            yield k
+        if self.parent:
+            for k in self.parent._keys():
+                if k not in yielded:
+                    yielded.add(k)
+                    yield k
+    def get(self, k, d=None):
+        try:
+            return self[k]
+        except KeyError:
+            return d
+    def get_color(self, k, d=None):
+        """
+        Strip leading # off colors if necessary
+        """
+        color = self.get(k, d)
+        if hasattr(color, 'startswith') and color.startswith('#'):
+            color = color[1:]
+            if len(color) == 3:  # Premailers reduces colors like #00ff00 to #0f0, openpyxl doesn't like that
+                color = ''.join(2 * c for c in color)
+        return color
+class Element(object):
+    """
+    Our base class for representing an html element along with a cascading style.
+    The element is created along with a parent so that the StyleDict that we store
+    can point to the parent's StyleDict.
+    """
+    def __init__(self, element, parent=None):
+        self.element = element
+        self.number_format = None
+        parent_style = parent.style_dict if parent else None
+        self.style_dict = StyleDict(style_string_to_dict(element.get('style', '')), parent=parent_style)
+        self._style_cache = None
+    def style(self):
+        """
+        Turn the css styles for this element into an openpyxl NamedStyle.
+        """
+        if not self._style_cache:
+            self._style_cache = style_dict_to_named_style(self.style_dict, number_format=self.number_format)
+        return self._style_cache
+    def get_dimension(self, dimension_key):
+        """
+        Extracts the dimension from the style dict of the Element and returns it as a float.
+        """
+        dimension = self.style_dict.get(dimension_key)
+        if dimension:
+            if dimension[-2:] in ['px', 'em', 'pt', 'in', 'cm']:
+                dimension = dimension[:-2]
+            dimension = float(dimension)
+        return dimension
+class Table(Element):
+    """
+    The concrete implementations of Elements are semantically named for the types of elements we are interested in.
+    This defines a very concrete tree structure for html tables that we expect to deal with. I prefer this compared to
+    allowing Element to have an arbitrary number of children and dealing with an abstract element tree.
+    """
+    def __init__(self, table):
+        """
+        takes an html table object (from lxml)
+        """
+        super(Table, self).__init__(table)
+        table_head = table.find('thead')
+        self.head = TableHead(table_head, parent=self) if table_head is not None else None
+        table_body = table.find('tbody')
+        self.body = TableBody(table_body if table_body is not None else table, parent=self)
+class TableHead(Element):
+    """
+    This class maps to the `<th>` element of the html table.
+    """
+    def __init__(self, head, parent=None):
+        super(TableHead, self).__init__(head, parent=parent)
+        self.rows = [TableRow(tr, parent=self) for tr in head.findall('tr')]
+class TableBody(Element):
+    """
+    This class maps to the `<tbody>` element of the html table.
+    """
+    def __init__(self, body, parent=None):
+        super(TableBody, self).__init__(body, parent=parent)
+        self.rows = [TableRow(tr, parent=self) for tr in body.findall('tr')]
+class TableRow(Element):
+    """
+    This class maps to the `<tr>` element of the html table.
+    """
+    def __init__(self, tr, parent=None):
+        super(TableRow, self).__init__(tr, parent=parent)
+        self.cells = [TableCell(cell, parent=self) for cell in tr.findall('th') + tr.findall('td')]
+def element_to_string(el):
+    return _element_to_string(el).strip()
+def _element_to_string(el):
+    string = ''
+    for x in el.iterchildren():
+        string += '\n' + _element_to_string(x)
+    text = el.text.strip() if el.text else ''
+    tail = el.tail.strip() if el.tail else ''
+    return text + string + '\n' + tail
+class TableCell(Element):
+    """
+    This class maps to the `<td>` element of the html table.
+    """
+    CELL_TYPES = {'TYPE_STRING', 'TYPE_FORMULA', 'TYPE_NUMERIC', 'TYPE_BOOL', 'TYPE_CURRENCY', 'TYPE_PERCENTAGE',
+                  'TYPE_NULL', 'TYPE_INLINE', 'TYPE_ERROR', 'TYPE_FORMULA_CACHE_STRING', 'TYPE_INTEGER'}
+    def __init__(self, cell, parent=None):
+        super(TableCell, self).__init__(cell, parent=parent)
+        self.value = element_to_string(cell)
+        self.number_format = self.get_number_format()
+    def data_type(self):
+        cell_types = self.CELL_TYPES & set(self.element.get('class', '').split())
+        if cell_types:
+            if 'TYPE_FORMULA' in cell_types:
+                # Make sure TYPE_FORMULA takes precedence over the other classes in the set.
+                cell_type = 'TYPE_FORMULA'
+            elif cell_types & {'TYPE_CURRENCY', 'TYPE_INTEGER', 'TYPE_PERCENTAGE'}:
+                cell_type = 'TYPE_NUMERIC'
+            else:
+                cell_type = cell_types.pop()
+        else:
+            cell_type = 'TYPE_STRING'
+        return getattr(cell, cell_type)
+    def get_number_format(self):
+        if 'TYPE_CURRENCY' in self.element.get('class', '').split():
+            return FORMAT_CURRENCY_USD_SIMPLE
+        if 'TYPE_INTEGER' in self.element.get('class', '').split():
+            return '#,##0'
+        if 'TYPE_PERCENTAGE' in self.element.get('class', '').split():
+            return FORMAT_PERCENTAGE
+        if 'TYPE_DATE' in self.element.get('class', '').split():
+            return FORMAT_DATE_MMDDYYYY
+        if self.data_type() == cell.TYPE_NUMERIC:
+            try:
+                int(self.value)
+            except ValueError:
+                return '#,##0.##'
+            else:
+                return '#,##0'
+    def format(self, cell):
+        cell.style = self.style()
+        data_type = self.data_type()
+        if data_type:
+            cell.data_type = data_type
\ No newline at end of file
--- a/ppstructure/table/tablepyxl/tablepyxl.py
+++ b/ppstructure/table/tablepyxl/tablepyxl.py
+# Do imports like python3 so our package works for 2 and 3
+from __future__ import absolute_import
+from lxml import html
+from openpyxl import Workbook
+from openpyxl.utils import get_column_letter
+from premailer import Premailer
+from tablepyxl.style import Table
+def string_to_int(s):
+    if s.isdigit():
+        return int(s)
+    return 0
+def get_Tables(doc):
+    tree = html.fromstring(doc)
+    comments = tree.xpath('//comment()')
+    for comment in comments:
+        comment.drop_tag()
+    return [Table(table) for table in tree.xpath('//table')]
+def write_rows(worksheet, elem, row, column=1):
+    """
+    Writes every tr child element of elem to a row in the worksheet
+    returns the next row after all rows are written
+    """
+    from openpyxl.cell.cell import MergedCell
+    initial_column = column
+    for table_row in elem.rows:
+        for table_cell in table_row.cells:
+            cell = worksheet.cell(row=row, column=column)
+            while isinstance(cell, MergedCell):
+                column += 1
+                cell = worksheet.cell(row=row, column=column)
+            colspan = string_to_int(table_cell.element.get("colspan", "1"))
+            rowspan = string_to_int(table_cell.element.get("rowspan", "1"))
+            if rowspan > 1 or colspan > 1:
+                worksheet.merge_cells(start_row=row, start_column=column,
+                                      end_row=row + rowspan - 1, end_column=column + colspan - 1)
+            cell.value = table_cell.value
+            table_cell.format(cell)
+            min_width = table_cell.get_dimension('min-width')
+            max_width = table_cell.get_dimension('max-width')
+            if colspan == 1:
+                # Initially, when iterating for the first time through the loop, the width of all the cells is None.
+                # As we start filling in contents, the initial width of the cell (which can be retrieved by:
+                # worksheet.column_dimensions[get_column_letter(column)].width) is equal to the width of the previous
+                # cell in the same column (i.e. width of A2 = width of A1)
+                width = max(worksheet.column_dimensions[get_column_letter(column)].width or 0, len(table_cell.value) + 2)
+                if max_width and width > max_width:
+                    width = max_width
+                elif min_width and width < min_width:
+                    width = min_width
+                worksheet.column_dimensions[get_column_letter(column)].width = width
+            column += colspan
+        row += 1
+        column = initial_column
+    return row
+def table_to_sheet(table, wb):
+    """
+    Takes a table and workbook and writes the table to a new sheet.
+    The sheet title will be the same as the table attribute name.
+    """
+    ws = wb.create_sheet(title=table.element.get('name'))
+    insert_table(table, ws, 1, 1)
+def document_to_workbook(doc, wb=None, base_url=None):
+    """
+    Takes a string representation of an html document and writes one sheet for
+    every table in the document.
+    The workbook is returned
+    """
+    if not wb:
+        wb = Workbook()
+        wb.remove(wb.active)
+    inline_styles_doc = Premailer(doc, base_url=base_url, remove_classes=False).transform()
+    tables = get_Tables(inline_styles_doc)
+    for table in tables:
+        table_to_sheet(table, wb)
+    return wb
+def document_to_xl(doc, filename, base_url=None):
+    """
+    Takes a string representation of an html document and writes one sheet for
+    every table in the document. The workbook is written out to a file called filename
+    """
+    wb = document_to_workbook(doc, base_url=base_url)
+    wb.save(filename)
+def insert_table(table, worksheet, column, row):
+    if table.head:
+        row = write_rows(worksheet, table.head, row, column)
+    if table.body:
+        row = write_rows(worksheet, table.body, row, column)
+def insert_table_at_cell(table, cell):
+    """
+    Inserts a table at the location of an openpyxl Cell object.
+    """
+    ws = cell.parent
+    column, row = cell.column, cell.row
+    insert_table(table, ws, column, row)
\ No newline at end of file
--- a/ppstructure/utility.py
+++ b/ppstructure/utility.py
+# copyright (c) 2020 PaddlePaddle Authors. All Rights Reserve.
+#
+# Licensed under the Apache License, Version 2.0 (the "License");
+# you may not use this file except in compliance with the License.
+# You may obtain a copy of the License at
+#
+#    http://www.apache.org/licenses/LICENSE-2.0
+#
+# Unless required by applicable law or agreed to in writing, software
+# distributed under the License is distributed on an "AS IS" BASIS,
+# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
+# See the License for the specific language governing permissions and
+# limitations under the License.
+from PIL import Image
+import numpy as np
+from tools.infer.utility import draw_ocr_box_txt, init_args as infer_args
+def init_args():
+    parser = infer_args()
+    # params for output
+    parser.add_argument("--output", type=str, default='./output/table')
+    # params for table structure
+    parser.add_argument("--table_max_len", type=int, default=488)
+    parser.add_argument("--table_model_dir", type=str)
+    parser.add_argument("--table_char_type", type=str, default='en')
+    parser.add_argument("--table_char_dict_path", type=str, default="../ppocr/utils/dict/table_structure_dict.txt")
+    parser.add_argument("--layout_path_model", type=str, default="lp://PubLayNet/ppyolov2_r50vd_dcn_365e_publaynet/config")
+    return parser
+def parse_args():
+    parser = init_args()
+    return parser.parse_args()
+def draw_structure_result(image, result, font_path):
+    if isinstance(image, np.ndarray):
+        image = Image.fromarray(image)
+    boxes, txts, scores = [], [], []
+    for region in result:
+        if region['type'] == 'Table':
+            pass
+        else:
+            for box, rec_res in zip(region['res'][0], region['res'][1]):
+                boxes.append(np.array(box).reshape(-1, 2))
+                txts.append(rec_res[0])
+                scores.append(rec_res[1])
+    im_show = draw_ocr_box_txt(image, boxes, txts, scores, font_path=font_path,drop_score=0)
+    return im_show
\ No newline at end of file
--- a/requirements.txt
+++ b/requirements.txt
@@ -7,4 +7,4 @@ tqdm
 numpy
 visualdl
 python-Levenshtein
-opencv-contrib-python==4.2.0.32
+opencv-contrib-python==4.4.0.46
\ No newline at end of file
--- a/setup.py
+++ b/setup.py
@@ -14,6 +14,7 @@
 from setuptools import setup
 from io import open
+from paddleocr import VERSION
 with open('requirements.txt', encoding="utf-8-sig") as f:
    requirements = f.readlines()
@@ -32,7 +33,7 @@ setup(
    package_dir={'paddleocr': ''},
    include_package_data=True,
    entry_points={"console_scripts": ["paddleocr= paddleocr.paddleocr:main"]},
-    version='2.0.6',
+    version=VERSION,
    install_requires=requirements,
    license='Apache License 2.0',
    description='Awesome OCR toolkits based on PaddlePaddle （8.6M ultra-lightweight pre-trained model, support training and deployment among server, mobile, embeded and IoT devices',

--- a/tests/ocr_det_params.txt
+++ b/tests/ocr_det_params.txt
+===========================train_params===========================
+model_name:ocr_det
+python:python3.7
+gpu_list:0|0,1
+Global.use_gpu:True|True
+Global.auto_cast:null
+Global.epoch_num:lite_train_infer=2|whole_train_infer=300
+Global.save_model_dir:./output/
+Train.loader.batch_size_per_card:lite_train_infer=2|whole_train_infer=4
+Global.pretrained_model:null
+train_model_name:latest
+train_infer_img_dir:./train_data/icdar2015/text_localization/ch4_test_images/
+null:null
+##
+trainer:norm_train|pact_train
+norm_train:tools/train.py -c configs/det/det_mv3_db.yml -o Global.pretrained_model=./pretrain_models/MobileNetV3_large_x0_5_pretrained
+pact_train:deploy/slim/quantization/quant.py -c configs/det/det_mv3_db.yml -o
+fpgm_train:null
+distill_train:null
+null:null
+null:null
+##
+===========================eval_params=========================== 
+eval:tools/eval.py -c configs/det/det_mv3_db.yml -o
+null:null
+##
+===========================infer_params===========================
+Global.save_inference_dir:./output/
+Global.pretrained_model:
+norm_export:tools/export_model.py -c configs/det/det_mv3_db.yml -o 
+quant_export:deploy/slim/quantization/export_model.py -c configs/det/det_mv3_db.yml -o 
+fpgm_export:deploy/slim/prune/export_prune_model.py
+distill_export:null
+export1:null
+export2:null
+##
+infer_model:./inference/ch_ppocr_mobile_v2.0_det_infer/
+infer_export:null
+infer_quant:False
+inference:tools/infer/predict_det.py
+--use_gpu:True|False
+--enable_mkldnn:True|False
+--cpu_threads:1|6
+--rec_batch_num:1
+--use_tensorrt:False|True
+--precision:fp32|fp16|int8
+--det_model_dir:
+--image_dir:./inference/ch_det_data_50/all-sum-510/
+--save_log_path:null
+--benchmark:True
+null:null
--- a/tests/ocr_rec_params.txt
+++ b/tests/ocr_rec_params.txt
+===========================train_params===========================
+model_name:ocr_rec
+python:python3.7
+gpu_list:0|2,3
+Global.use_gpu:True|True
+Global.auto_cast:null
+Global.epoch_num:lite_train_infer=2|whole_train_infer=300
+Global.save_model_dir:./output/
+Train.loader.batch_size_per_card:lite_train_infer=128|whole_train_infer=128
+Global.pretrained_model:null
+train_model_name:latest
+train_infer_img_dir:./train_data/ic15_data/train
+null:null
+##
+trainer:norm_train|pact_train
+norm_train:tools/train.py -c configs/rec/rec_icdar15_train.yml -o
+pact_train:deploy/slim/quantization/quant.py -c configs/rec/rec_icdar15_train.yml -o
+fpgm_train:null
+distill_train:null
+null:null
+null:null
+##
+===========================eval_params===========================
+eval:tools/eval.py -c configs/rec/rec_icdar15_train.yml -o
+null:null
+##
+===========================infer_params===========================
+Global.save_inference_dir:./output/
+Global.pretrained_model:
+norm_export:tools/export_model.py -c configs/rec/rec_icdar15_train.yml -o
+quant_export:deploy/slim/quantization/export_model.py -c configs/rec/rec_icdar15_train.yml -o
+fpgm_export:null
+distill_export:null
+export1:null
+export2:null
+##
+infer_model:./inference/ch_ppocr_mobile_v2.0_rec_infer/
+infer_export:null
+infer_quant:False
+inference:tools/infer/predict_rec.py
+--use_gpu:True|False
+--enable_mkldnn:True|False
+--cpu_threads:1|6
+--rec_batch_num:1
+--use_tensorrt:True|False
+--precision:fp32|fp16|int8
+--rec_model_dir:
+--image_dir:./inference/rec_inference
+--save_log_path:./test/output/
+--benchmark:True
+null:null
--- a/tests/prepare.sh
+++ b/tests/prepare.sh
+#!/bin/bash
+FILENAME=$1
+# MODE be one of ['lite_train_infer' 'whole_infer' 'whole_train_infer', 'infer']
+MODE=$2
+dataline=$(cat ${FILENAME})
+# parser params
+IFS=$'\n'
+lines=(${dataline})
+function func_parser_key(){
+    strs=$1
+    IFS=":"
+    array=(${strs})
+    tmp=${array[0]}
+    echo ${tmp}
+}
+function func_parser_value(){
+    strs=$1
+    IFS=":"
+    array=(${strs})
+    tmp=${array[1]}
+    echo ${tmp}
+}
+IFS=$'\n'
+# The training params
+model_name=$(func_parser_value "${lines[1]}")
+trainer_list=$(func_parser_value "${lines[14]}")
+# MODE be one of ['lite_train_infer' 'whole_infer' 'whole_train_infer']
+MODE=$2
+if [ ${MODE} = "lite_train_infer" ];then
+    # pretrain lite train data
+    wget -nc -P  ./pretrain_models/ https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV3_large_x0_5_pretrained.pdparams
+    rm -rf ./train_data/icdar2015
+    rm -rf ./train_data/ic15_data
+    wget -nc -P ./train_data/ https://paddleocr.bj.bcebos.com/dygraph_v2.0/test/icdar2015_lite.tar
+    wget -nc -P ./train_data/ https://paddleocr.bj.bcebos.com/dygraph_v2.0/test/ic15_data.tar # todo change to bcebos
+    cd ./train_data/ && tar xf icdar2015_lite.tar && tar xf ic15_data.tar
+    ln -s ./icdar2015_lite ./icdar2015
+    cd ../
+elif [ ${MODE} = "whole_train_infer" ];then
+    wget -nc -P  ./pretrain_models/ https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV3_large_x0_5_pretrained.pdparams
+    rm -rf ./train_data/icdar2015
+    rm -rf ./train_data/ic15_data
+    wget -nc -P ./train_data/ https://paddleocr.bj.bcebos.com/dygraph_v2.0/test/icdar2015.tar
+    wget -nc -P ./train_data/ https://paddleocr.bj.bcebos.com/dygraph_v2.0/test/ic15_data.tar
+    cd ./train_data/ && tar xf icdar2015.tar && tar xf ic15_data.tar && cd ../
+elif [ ${MODE} = "whole_infer" ];then
+    wget -nc -P  ./pretrain_models/ https://paddle-imagenet-models-name.bj.bcebos.com/dygraph/MobileNetV3_large_x0_5_pretrained.pdparams
+    rm -rf ./train_data/icdar2015
+    rm -rf ./train_data/ic15_data
+    wget -nc -P ./train_data/ https://paddleocr.bj.bcebos.com/dygraph_v2.0/test/icdar2015_infer.tar
+    wget -nc -P ./train_data/ https://paddleocr.bj.bcebos.com/dygraph_v2.0/test/ic15_data.tar
+    cd ./train_data/ && tar xf icdar2015_infer.tar && tar xf ic15_data.tar
+    ln -s ./icdar2015_infer ./icdar2015
+    cd ../
+else
+    if [ ${model_name} = "ocr_det" ]; then
+        eval_model_name="ch_ppocr_mobile_v2.0_det_infer"
+        rm -rf ./train_data/icdar2015
+        wget -nc -P ./inference https://paddleocr.bj.bcebos.com/dygraph_v2.0/test/ch_det_data_50.tar
+        wget -nc  -P ./inference https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_det_infer.tar
+        cd ./inference && tar xf ${eval_model_name}.tar && tar xf ch_det_data_50.tar && cd ../
+    else 
+        rm -rf ./train_data/ic15_data
+        eval_model_name="ch_ppocr_mobile_v2.0_rec_infer"
+        wget -nc -P ./train_data/ https://paddleocr.bj.bcebos.com/dygraph_v2.0/test/ic15_data.tar
+        wget -nc  -P ./inference https://paddleocr.bj.bcebos.com/dygraph_v2.0/ch/ch_ppocr_mobile_v2.0_rec_infer.tar
+        cd ./inference && tar xf ${eval_model_name}.tar && tar xf ic15_data.tar && cd ../
+    fi 
+fi
--- a/tests/readme.md
+++ b/tests/readme.md
+# 介绍
+test.sh和params.txt文件配合使用，完成OCR轻量检测和识别模型从训练到预测的流程测试。
+# 安装依赖
+- 安装PaddlePaddle >= 2.0
+- 安装PaddleOCR依赖
+    ```
+    pip3 install  -r ../requirements.txt
+    ```
+- 安装autolog
+    ```
+    git clone https://github.com/LDOUBLEV/AutoLog
+    cd AutoLog
+    pip3 install -r requirements.txt
+    python3 setup.py bdist_wheel
+    pip3 install ./dist/auto_log-1.0.0-py3-none-any.whl
+    cd ../
+    ```
+# 目录介绍
+```bash
+tests/
+├── ocr_det_params.txt   # 测试OCR检测模型的参数配置文件
+├── ocr_rec_params.txt   # 测试OCR识别模型的参数配置文件
+└── prepare.sh           # 完成test.sh运行所需要的数据和模型下载
+└── test.sh              # 根据
+```
+# 使用方法
+test.sh包含四种运行模式，每种模式的运行数据不同，分别用于测试速度和精度，分别是：
+- 模式1 lite_train_infer，使用少量数据训练，用于快速验证训练到预测的走通流程，不验证精度和速度；
+```
+bash test/prepare.sh ./tests/ocr_det_params.txt 'lite_train_infer'
+bash tests/test.sh ./tests/ocr_det_params.txt 'lite_train_infer'
+```
+- 模式2 whole_infer，使用少量数据训练，一定量数据预测，用于验证训练后的模型执行预测，预测速度是否合理；
+```
+bash tests/prepare.sh ./tests/ocr_det_params.txt 'whole_infer'
+bash tests/test.sh ./tests/ocr_det_params.txt 'whole_infer'
+```
+- 模式3 infer 不训练，全量数据预测，走通开源模型评估、动转静，检查inference model预测时间和精度;
+```
+bash tests/prepare.sh ./tests/ocr_det_params.txt 'infer'
+用法1:
+bash tests/test.sh ./tests/ocr_det_params.txt 'infer'
+用法2: 指定GPU卡预测，第三个传入参数为GPU卡号
+bash tests/test.sh ./tests/ocr_det_params.txt 'infer' '1'
+```
+模式4: whole_train_infer , CE： 全量数据训练，全量数据预测，验证模型训练精度，预测精度，预测速度
+```
+bash tests/prepare.sh ./tests/ocr_det_params.txt 'whole_train_infer'
+bash tests/test.sh ./tests/ocr_det_params.txt 'whole_train_infer'
+```
--- a/tests/test.sh
+++ b/tests/test.sh
+#!/bin/bash
+FILENAME=$1
+# MODE be one of ['lite_train_infer' 'whole_infer' 'whole_train_infer', 'infer']
+MODE=$2
+dataline=$(cat ${FILENAME})
+# parser params
+IFS=$'\n'
+lines=(${dataline})
+function func_parser_key(){
+    strs=$1
+    IFS=":"
+    array=(${strs})
+    tmp=${array[0]}
+    echo ${tmp}
+}
+function func_parser_value(){
+    strs=$1
+    IFS=":"
+    array=(${strs})
+    tmp=${array[1]}
+    echo ${tmp}
+}
+function func_set_params(){
+    key=$1
+    value=$2
+    if [ ${key} = "null" ];then
+        echo " "
+    elif [[ ${value} = "null" ]] || [[ ${value} = " " ]] || [ ${#value} -le 0 ];then
+        echo " "
+    else 
+        echo "${key}=${value}"
+    fi
+}
+function func_parser_params(){
+    strs=$1
+    IFS=":"
+    array=(${strs})
+    key=${array[0]}
+    tmp=${array[1]}
+    IFS="|"
+    res=""
+    for _params in ${tmp[*]}; do
+        IFS="="
+        array=(${_params})
+        mode=${array[0]}
+        value=${array[1]}
+        if [[ ${mode} = ${MODE} ]]; then
+            IFS="|"
+            #echo $(func_set_params "${mode}" "${value}")
+            echo $value
+            break
+        fi
+        IFS="|"
+    done
+    echo ${res}
+}
+function status_check(){
+    last_status=$1   # the exit code
+    run_command=$2
+    run_log=$3
+    if [ $last_status -eq 0 ]; then
+        echo -e "\033[33m Run successfully with command - ${run_command}!  \033[0m" | tee -a ${run_log}
+    else
+        echo -e "\033[33m Run failed with command - ${run_command}!  \033[0m" | tee -a ${run_log}
+    fi
+}
+IFS=$'\n'
+# The training params
+model_name=$(func_parser_value "${lines[1]}")
+python=$(func_parser_value "${lines[2]}")
+gpu_list=$(func_parser_value "${lines[3]}")
+train_use_gpu_key=$(func_parser_key "${lines[4]}")
+train_use_gpu_value=$(func_parser_value "${lines[4]}")
+autocast_list=$(func_parser_value "${lines[5]}")
+autocast_key=$(func_parser_key "${lines[5]}")
+epoch_key=$(func_parser_key "${lines[6]}")
+epoch_num=$(func_parser_params "${lines[6]}")
+save_model_key=$(func_parser_key "${lines[7]}")
+train_batch_key=$(func_parser_key "${lines[8]}")
+train_batch_value=$(func_parser_params "${lines[8]}")
+pretrain_model_key=$(func_parser_key "${lines[9]}")
+pretrain_model_value=$(func_parser_value "${lines[9]}")
+train_model_name=$(func_parser_value "${lines[10]}")
+train_infer_img_dir=$(func_parser_value "${lines[11]}")
+train_param_key1=$(func_parser_key "${lines[12]}")
+train_param_value1=$(func_parser_value "${lines[12]}")
+trainer_list=$(func_parser_value "${lines[14]}")
+trainer_norm=$(func_parser_key "${lines[15]}")
+norm_trainer=$(func_parser_value "${lines[15]}")
+pact_key=$(func_parser_key "${lines[16]}")
+pact_trainer=$(func_parser_value "${lines[16]}")
+fpgm_key=$(func_parser_key "${lines[17]}")
+fpgm_trainer=$(func_parser_value "${lines[17]}")
+distill_key=$(func_parser_key "${lines[18]}")
+distill_trainer=$(func_parser_value "${lines[18]}")
+trainer_key1=$(func_parser_key "${lines[19]}")
+trainer_value1=$(func_parser_value "${lines[19]}")
+trainer_key2=$(func_parser_key "${lines[20]}")
+trainer_value2=$(func_parser_value "${lines[20]}")
+eval_py=$(func_parser_value "${lines[23]}")
+eval_key1=$(func_parser_key "${lines[24]}")
+eval_value1=$(func_parser_value "${lines[24]}")
+save_infer_key=$(func_parser_key "${lines[27]}")
+export_weight=$(func_parser_key "${lines[28]}")
+norm_export=$(func_parser_value "${lines[29]}")
+pact_export=$(func_parser_value "${lines[30]}")
+fpgm_export=$(func_parser_value "${lines[31]}")
+distill_export=$(func_parser_value "${lines[32]}")
+export_key1=$(func_parser_key "${lines[33]}")
+export_value1=$(func_parser_value "${lines[33]}")
+export_key2=$(func_parser_key "${lines[34]}")
+export_value2=$(func_parser_value "${lines[34]}")
+# parser inference model 
+infer_model_dir_list=$(func_parser_value "${lines[36]}")
+infer_export_list=$(func_parser_value "${lines[37]}")
+infer_is_quant=$(func_parser_value "${lines[38]}")
+# parser inference 
+inference_py=$(func_parser_value "${lines[39]}")
+use_gpu_key=$(func_parser_key "${lines[40]}")
+use_gpu_list=$(func_parser_value "${lines[40]}")
+use_mkldnn_key=$(func_parser_key "${lines[41]}")
+use_mkldnn_list=$(func_parser_value "${lines[41]}")
+cpu_threads_key=$(func_parser_key "${lines[42]}")
+cpu_threads_list=$(func_parser_value "${lines[42]}")
+batch_size_key=$(func_parser_key "${lines[43]}")
+batch_size_list=$(func_parser_value "${lines[43]}")
+use_trt_key=$(func_parser_key "${lines[44]}")
+use_trt_list=$(func_parser_value "${lines[44]}")
+precision_key=$(func_parser_key "${lines[45]}")
+precision_list=$(func_parser_value "${lines[45]}")
+infer_model_key=$(func_parser_key "${lines[46]}")
+image_dir_key=$(func_parser_key "${lines[47]}")
+infer_img_dir=$(func_parser_value "${lines[47]}")
+save_log_key=$(func_parser_key "${lines[48]}")
+benchmark_key=$(func_parser_key "${lines[49]}")
+benchmark_value=$(func_parser_value "${lines[49]}")
+infer_key1=$(func_parser_key "${lines[50]}")
+infer_value1=$(func_parser_value "${lines[50]}")
+LOG_PATH="./tests/output"
+mkdir -p ${LOG_PATH}
+status_log="${LOG_PATH}/results.log"
+function func_inference(){
+    IFS='|'
+    _python=$1
+    _script=$2
+    _model_dir=$3
+    _log_path=$4
+    _img_dir=$5
+    _flag_quant=$6
+    # inference 
+    for use_gpu in ${use_gpu_list[*]}; do
+        if [ ${use_gpu} = "False" ] || [ ${use_gpu} = "cpu" ]; then
+            for use_mkldnn in ${use_mkldnn_list[*]}; do
+                if [ ${use_mkldnn} = "False" ] && [ ${_flag_quant} = "True" ]; then
+                    continue
+                fi
+                for threads in ${cpu_threads_list[*]}; do
+                    for batch_size in ${batch_size_list[*]}; do
+                        _save_log_path="${_log_path}/infer_cpu_usemkldnn_${use_mkldnn}_threads_${threads}_batchsize_${batch_size}.log"
+                        set_infer_data=$(func_set_params "${image_dir_key}" "${_img_dir}")
+                        set_benchmark=$(func_set_params "${benchmark_key}" "${benchmark_value}")
+                        set_batchsize=$(func_set_params "${batch_size_key}" "${batch_size}")
+                        set_cpu_threads=$(func_set_params "${cpu_threads_key}" "${threads}")
+                        set_model_dir=$(func_set_params "${infer_model_key}" "${_model_dir}")
+                        set_infer_params1=$(func_set_params "${infer_key1}" "${infer_value1}")
+                        command="${_python} ${_script} ${use_gpu_key}=${use_gpu} ${use_mkldnn_key}=${use_mkldnn} ${set_cpu_threads} ${set_model_dir} ${set_batchsize} ${set_infer_data} ${set_benchmark} ${set_infer_params1} > ${_save_log_path} 2>&1 "
+                        eval $command
+                        last_status=${PIPESTATUS[0]}
+                        eval "cat ${_save_log_path}"
+                        status_check $last_status "${command}" "${status_log}"
+                    done
+                done
+            done
+        elif [ ${use_gpu} = "True" ] || [ ${use_gpu} = "gpu" ]; then
+            for use_trt in ${use_trt_list[*]}; do
+                for precision in ${precision_list[*]}; do
+                    if [[ ${_flag_quant} = "False" ]] && [[ ${precision} =~ "int8" ]]; then
+                        continue
+                    fi 
+                    if [[ ${precision} =~ "fp16" || ${precision} =~ "int8" ]] && [ ${use_trt} = "False" ]; then
+                        continue
+                    fi
+                    if [[ ${use_trt} = "False" || ${precision} =~ "int8" ]] && [ ${_flag_quant} = "True" ]; then
+                        continue
+                    fi
+                    for batch_size in ${batch_size_list[*]}; do
+                        _save_log_path="${_log_path}/infer_gpu_usetrt_${use_trt}_precision_${precision}_batchsize_${batch_size}.log"
+                        set_infer_data=$(func_set_params "${image_dir_key}" "${_img_dir}")
+                        set_benchmark=$(func_set_params "${benchmark_key}" "${benchmark_value}")
+                        set_batchsize=$(func_set_params "${batch_size_key}" "${batch_size}")
+                        set_tensorrt=$(func_set_params "${use_trt_key}" "${use_trt}")
+                        set_precision=$(func_set_params "${precision_key}" "${precision}")
+                        set_model_dir=$(func_set_params "${infer_model_key}" "${_model_dir}")
+                        set_infer_params1=$(func_set_params "${infer_key1}" "${infer_value1}")
+                        command="${_python} ${_script} ${use_gpu_key}=${use_gpu} ${set_tensorrt} ${set_precision} ${set_model_dir} ${set_batchsize} ${set_infer_data} ${set_benchmark} ${set_infer_params1} > ${_save_log_path} 2>&1 "
+                        eval $command
+                        last_status=${PIPESTATUS[0]}
+                        eval "cat ${_save_log_path}"
+                        status_check $last_status "${command}" "${status_log}"
+                    done
+                done
+            done
+        else
+            echo "Does not support hardware other than CPU and GPU Currently!"
+        fi
+    done
+}
+if [ ${MODE} = "infer" ]; then
+    GPUID=$3
+    if [ ${#GPUID} -le 0 ];then
+        env=" "
+    else
+        env="export CUDA_VISIBLE_DEVICES=${GPUID}"
+    fi
+    # set CUDA_VISIBLE_DEVICES
+    eval $env
+    export Count=0
+    IFS="|"
+    infer_run_exports=(${infer_export_list})
+    infer_quant_flag=(${infer_is_quant})
+    for infer_model in ${infer_model_dir_list[*]}; do
+        # run export
+        if [ ${infer_run_exports[Count]} != "null" ];then
+            save_infer_dir=$(dirname $infer_model)
+            set_export_weight=$(func_set_params "${export_weight}" "${infer_model}")
+            set_save_infer_key=$(func_set_params "${save_infer_key}" "${save_infer_dir}")
+            export_cmd="${python} ${norm_export} ${set_export_weight} ${set_save_infer_key}"
+            eval $export_cmd
+            status_export=$?
+            if [ ${status_export} = 0 ];then
+                status_check $status_export "${export_cmd}" "${status_log}"
+            fi
+        else
+            save_infer_dir=${infer_model}
+        fi
+        #run inference
+        is_quant=${infer_quant_flag[Count]}
+        func_inference "${python}" "${inference_py}" "${save_infer_dir}" "${LOG_PATH}" "${infer_img_dir}" ${is_quant}
+        Count=$(($Count + 1))
+    done
+else
+    IFS="|"
+    export Count=0
+    USE_GPU_KEY=(${train_use_gpu_value})
+    for gpu in ${gpu_list[*]}; do
+        use_gpu=${USE_GPU_KEY[Count]}
+        Count=$(($Count + 1))
+        if [ ${gpu} = "-1" ];then
+            env=""
+        elif [ ${#gpu} -le 1 ];then
+            env="export CUDA_VISIBLE_DEVICES=${gpu}"
+            eval ${env}
+        elif [ ${#gpu} -le 15 ];then
+            IFS=","
+            array=(${gpu})
+            env="export CUDA_VISIBLE_DEVICES=${array[0]}"
+            IFS="|"
+        else
+            IFS=";"
+            array=(${gpu})
+            ips=${array[0]}
+            gpu=${array[1]}
+            IFS="|"
+            env=" "
+        fi
+        for autocast in ${autocast_list[*]}; do 
+            for trainer in ${trainer_list[*]}; do 
+                flag_quant=False
+                if [ ${trainer} = ${pact_key} ]; then
+                    run_train=${pact_trainer}
+                    run_export=${pact_export}
+                    flag_quant=True
+                elif [ ${trainer} = "${fpgm_key}" ]; then
+                    run_train=${fpgm_trainer}
+                    run_export=${fpgm_export}
+                elif [ ${trainer} = "${distill_key}" ]; then
+                    run_train=${distill_trainer}
+                    run_export=${distill_export}
+                elif [ ${trainer} = ${trainer_key1} ]; then
+                    run_train=${trainer_value1}
+                    run_export=${export_value1}
+                elif [[ ${trainer} = ${trainer_key2} ]]; then
+                    run_train=${trainer_value2}
+                    run_export=${export_value2}
+                else
+                    run_train=${norm_trainer}
+                    run_export=${norm_export}
+                fi
+                if [ ${run_train} = "null" ]; then
+                    continue
+                fi
+                set_autocast=$(func_set_params "${autocast_key}" "${autocast}")
+                set_epoch=$(func_set_params "${epoch_key}" "${epoch_num}")
+                set_pretrain=$(func_set_params "${pretrain_model_key}" "${pretrain_model_value}")
+                set_batchsize=$(func_set_params "${train_batch_key}" "${train_batch_value}")
+                set_train_params1=$(func_set_params "${train_param_key1}" "${train_param_value1}")
+                set_use_gpu=$(func_set_params "${train_use_gpu_key}" "${use_gpu}")
+                save_log="${LOG_PATH}/${trainer}_gpus_${gpu}_autocast_${autocast}"
+                # load pretrain from norm training if current trainer is pact or fpgm trainer
+                if [ ${trainer} = ${pact_key} ] || [ ${trainer} = ${fpgm_key} ]; then
+                    set_pretrain="${load_norm_train_model}"
+                fi
+                set_save_model=$(func_set_params "${save_model_key}" "${save_log}")
+                if [ ${#gpu} -le 2 ];then  # train with cpu or single gpu
+                    cmd="${python} ${run_train} ${set_use_gpu}  ${set_save_model} ${set_epoch} ${set_pretrain} ${set_autocast} ${set_batchsize} ${set_train_params1} "
+                elif [ ${#gpu} -le 15 ];then  # train with multi-gpu
+                    cmd="${python} -m paddle.distributed.launch --gpus=${gpu} ${run_train} ${set_save_model} ${set_epoch} ${set_pretrain} ${set_autocast} ${set_batchsize} ${set_train_params1}"
+                else     # train with multi-machine
+                    cmd="${python} -m paddle.distributed.launch --ips=${ips} --gpus=${gpu} ${run_train} ${set_save_model} ${set_pretrain} ${set_epoch} ${set_autocast} ${set_batchsize} ${set_train_params1}"
+                fi
+                # run train
+                eval "unset CUDA_VISIBLE_DEVICES"
+                eval $cmd
+                status_check $? "${cmd}" "${status_log}"
+                set_eval_pretrain=$(func_set_params "${pretrain_model_key}" "${save_log}/${train_model_name}")
+                # save norm trained models to set pretrain for pact training and fpgm training 
+                if [ ${trainer} = ${trainer_norm} ]; then
+                    load_norm_train_model=${set_eval_pretrain}
+                fi
+                # run eval 
+                if [ ${eval_py} != "null" ]; then
+                    set_eval_params1=$(func_set_params "${eval_key1}" "${eval_value1}")
+                    eval_cmd="${python} ${eval_py} ${set_eval_pretrain} ${set_use_gpu} ${set_eval_params1}" 
+                    eval $eval_cmd
+                    status_check $? "${eval_cmd}" "${status_log}"
+                fi
+                # run export model
+                if [ ${run_export} != "null" ]; then 
+                    # run export model
+                    save_infer_path="${save_log}"
+                    set_export_weight=$(func_set_params "${export_weight}" "${save_log}/${train_model_name}")
+                    set_save_infer_key=$(func_set_params "${save_infer_key}" "${save_infer_path}")
+                    export_cmd="${python} ${run_export} ${set_export_weight} ${set_save_infer_key}"
+                    eval $export_cmd
+                    status_check $? "${export_cmd}" "${status_log}"
+                    #run inference
+                    eval $env
+                    save_infer_path="${save_log}"
+                    func_inference "${python}" "${inference_py}" "${save_infer_path}" "${LOG_PATH}" "${train_infer_img_dir}" "${flag_quant}"
+                    eval "unset CUDA_VISIBLE_DEVICES"
+                fi
+            done  # done with:    for trainer in ${trainer_list[*]}; do 
+        done      # done with:    for autocast in ${autocast_list[*]}; do 
+    done          # done with:    for gpu in ${gpu_list[*]}; do
+fi  # end if [ ${MODE} = "infer" ]; then