"test/vscode:/vscode.git/clone" did not exist on "4d4cdb3fe7dfbb87e5f138fe78954c6e1957b007"
README_ch.md 4.87 KB
Newer Older
WenmuZhou's avatar
WenmuZhou committed
1
2
# PaddleStructure

WenmuZhou's avatar
opt doc  
WenmuZhou committed
3
PaddleStructure是一个用于复杂版面分析的OCR工具包,其能够对图片形式的文档数据划分**文字、表格、标题、图片以及列表**5类区域,并将表格区域提取为excel
WenmuZhou's avatar
WenmuZhou committed
4

WenmuZhou's avatar
opt doc  
WenmuZhou committed
5
## 1. 快速开始
WenmuZhou's avatar
WenmuZhou committed
6

WenmuZhou's avatar
opt doc  
WenmuZhou committed
7
### 1.1 安装
WenmuZhou's avatar
WenmuZhou committed
8

WenmuZhou's avatar
opt doc  
WenmuZhou committed
9
10
**安装 layoutparser**
```sh
WenmuZhou's avatar
WenmuZhou committed
11
pip3 install -U https://paddleocr.bj.bcebos.com/whl/layoutparser-0.0.0-py3-none-any.whl
WenmuZhou's avatar
opt doc  
WenmuZhou committed
12
13
```
**安装 paddlestructure**
WenmuZhou's avatar
WenmuZhou committed
14

WenmuZhou's avatar
opt doc  
WenmuZhou committed
15
16
17
18
pip安装
```bash
pip install paddlestructure
```
WenmuZhou's avatar
WenmuZhou committed
19

WenmuZhou's avatar
opt doc  
WenmuZhou committed
20
21
22
23
24
本地构建并安装
```bash
python3 setup.py bdist_wheel
pip3 install dist/paddlestructure-x.x.x-py3-none-any.whl # x.x.x是 paddlestructure 的版本号
```
WenmuZhou's avatar
WenmuZhou committed
25

WenmuZhou's avatar
opt doc  
WenmuZhou committed
26
### 1.2 PaddleStructure whl包使用
WenmuZhou's avatar
WenmuZhou committed
27

WenmuZhou's avatar
opt doc  
WenmuZhou committed
28
#### 1.2.1 命令行使用
WenmuZhou's avatar
WenmuZhou committed
29

WenmuZhou's avatar
opt doc  
WenmuZhou committed
30
31
```bash
paddlestructure --image_dir=../doc/table/1.png
WenmuZhou's avatar
opt doc  
WenmuZhou committed
32
33
```

WenmuZhou's avatar
opt doc  
WenmuZhou committed
34
#### 1.2.2 Python脚本使用
WenmuZhou's avatar
WenmuZhou committed
35
36

```python
WenmuZhou's avatar
WenmuZhou committed
37
import os
WenmuZhou's avatar
WenmuZhou committed
38
import cv2
WenmuZhou's avatar
WenmuZhou committed
39
from paddlestructure import PaddleStructure,draw_result,save_res
WenmuZhou's avatar
WenmuZhou committed
40

WenmuZhou's avatar
WenmuZhou committed
41
table_engine = PaddleStructure(show_log=True)
WenmuZhou's avatar
WenmuZhou committed
42

WenmuZhou's avatar
WenmuZhou committed
43
save_folder = './output/table'
WenmuZhou's avatar
WenmuZhou committed
44
45
46
img_path = '../doc/table/1.png'
img = cv2.imread(img_path)
result = table_engine(img)
WenmuZhou's avatar
WenmuZhou committed
47
48
save_res(result, save_folder,os.path.basename(img_path).split('.')[0])

WenmuZhou's avatar
WenmuZhou committed
49
50
51
52
53
for line in result:
    print(line)

from PIL import Image

WenmuZhou's avatar
opt doc  
WenmuZhou committed
54
font_path = '../doc/fonts/simfang.ttf' # PaddleOCR下提供字体包
WenmuZhou's avatar
WenmuZhou committed
55
56
57
58
59
60
61
image = Image.open(img_path).convert('RGB')
im_show = draw_result(image, result,font_path=font_path)
im_show = Image.fromarray(im_show)
im_show.save('result.jpg')
```


WenmuZhou's avatar
opt doc  
WenmuZhou committed
62
63
64
65
66
67
68
69
70
#### 1.2.3 参数说明

| 字段            | 说明                                     | 默认值                                      |
| --------------- | ---------------------------------------- | ------------------------------------------- |
| output          | excel和识别结果保存的地址                | ./output/table                              |
| table_max_len   | 表格结构模型预测时,图像的长边resize尺度 | 488                                         |
| table_model_dir | 表格结构模型 inference 模型地址          | None                                        |
| table_char_type | 表格结构模型所用字典地址                 | ../ppocr/utils/dict/table_structure_dict.tx |

WenmuZhou's avatar
WenmuZhou committed
71
72
大部分参数和paddleocr whl包保持一致,见 [whl包文档](../doc/doc_ch/whl.md)

WenmuZhou's avatar
opt doc  
WenmuZhou committed
73
74
75
76
77
78
79
80
81
82
运行完成后,每张图片会在`output`字段指定的目录下有一个同名目录,图片里的每个表格会存储为一个excel,excel文件名为表格在图片里的坐标。


## 2. PaddleStructure Pipeline

流程如下
![pipeline](../doc/table/pipeline.jpg)

在PaddleStructure中,图片会先经由layoutparser进行版面分析,在版面分析中,会对图片里的区域进行分类,包括**文字、标题、图片、列表和表格**5类。对于前4类区域,直接使用PP-OCR完成对应区域文字检测与识别。对于表格类区域,经过Table OCR处理后,表格图片转换为相同表格样式的Excel文件。

WenmuZhou's avatar
WenmuZhou committed
83
### 2.1 版面分析
WenmuZhou's avatar
opt doc  
WenmuZhou committed
84
85
86

版面分析对文档数据进行区域分类,其中包括版面分析工具的Python脚本使用、提取指定类别检测框、性能指标以及自定义训练版面分析模型,详细内容可以参考[文档](layout/README.md)

WenmuZhou's avatar
WenmuZhou committed
87
### 2.2 表格识别
WenmuZhou's avatar
opt doc  
WenmuZhou committed
88
89
90

Table OCR将表格图片转换为excel文档,其中包含对于表格文本的检测和识别以及对于表格结构和单元格坐标的预测,详细说明参考[文档](table/README_ch.md)

WenmuZhou's avatar
WenmuZhou committed
91
## 3. 预测引擎推理
WenmuZhou's avatar
opt doc  
WenmuZhou committed
92
93
94
95

使用如下命令即可完成预测引擎的推理

```python
WenmuZhou's avatar
WenmuZhou committed
96
python3 table/predict_system.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 --vis_font_path=../doc/fonts/simfang.ttf
WenmuZhou's avatar
opt doc  
WenmuZhou committed
97
98
99
```
运行完成后,每张图片会output字段指定的目录下有一个同名目录,图片里的每个表格会存储为一个excel,excel文件名为表格在图片里的坐标。

WenmuZhou's avatar
WenmuZhou committed
100
**Model List**
WenmuZhou's avatar
WenmuZhou committed
101

WenmuZhou's avatar
opt doc  
WenmuZhou committed
102
103
104
|模型名称|模型简介|配置文件|推理模型大小|下载地址|
| --- | --- | --- | --- | --- |
|en_ppocr_mobile_v2.0_table_det|英文表格场景的文字检测|[ch_det_mv3_db_v2.0.yml](../configs/det/ch_ppocr_v2.0/ch_det_mv3_db_v2.0.yml)| 4.7M |[推理模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/table/en_ppocr_mobile_v2.0_table_det_infer.tar) |
WenmuZhou's avatar
del /  
WenmuZhou committed
105
|en_ppocr_mobile_v2.0_table_rec|英文表格场景的文字识别|[rec_chinese_lite_train_v2.0.yml](../configs/rec/rec_mv3_none_bilstm_ctc.yml)|6.9M|[推理模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/table/en_ppocr_mobile_v2.0_table_rec_infer.tar) |
WenmuZhou's avatar
opt doc  
WenmuZhou committed
106
|en_ppocr_mobile_v2.0_table_structure|英文表格场景的表格结构预测|[table_mv3.yml](../configs/table/table_mv3.yml)|18.6M|[推理模型](https://paddleocr.bj.bcebos.com/dygraph_v2.0/table/en_ppocr_mobile_v2.0_table_structure_infer.tar) |