README.md 4.15 KB
Newer Older
LDOUBLEV's avatar
LDOUBLEV committed
1
2

# 简介
LDOUBLEV's avatar
LDOUBLEV committed
3
PaddleOCR旨在打造一套丰富、领先、且实用的OCR工具库,助力使用者训练出更好的模型,并应用落地。
LDOUBLEV's avatar
LDOUBLEV committed
4
5
6

## 特性:
- 超轻量级模型
LDOUBLEV's avatar
LDOUBLEV committed
7
    - (检测模型4.1M + 识别模型4.5M = 8.6M)
LDOUBLEV's avatar
LDOUBLEV committed
8
- 支持竖排文字识别
LDOUBLEV's avatar
LDOUBLEV committed
9
10
11
12
13
14
    - (单模型同时支持横排和竖排文字识别)
- 支持长文本识别
- 支持中英文数字组合识别
- 提供训练代码
- 支持模型部署

LDOUBLEV's avatar
LDOUBLEV committed
15
16
17
18
19
20
21
22
23

## 文档教程
- [快速安装](./doc/installation.md)
- [快速开始]()
- [文本识别模型训练/评估/预测](./doc/detection.md)
- [文本预测模型训练/评估/预测](./doc/recognition.md)
- [基于inference model预测](./doc/)


LDOUBLEV's avatar
LDOUBLEV committed
24
25
## 文本检测算法:

LDOUBLEV's avatar
LDOUBLEV committed
26
PaddleOCR开源的文本检测算法列表:
LDOUBLEV's avatar
LDOUBLEV committed
27
28
29
30
31
- [x] [EAST](https://arxiv.org/abs/1704.03155)
- [x] [DB](https://arxiv.org/abs/1911.08947)
- [x] [SAST](https://arxiv.org/abs/1908.05498)
- []

LDOUBLEV's avatar
LDOUBLEV committed
32
33

算法效果:
LDOUBLEV's avatar
LDOUBLEV committed
34
35
36
37
38
39
|模型|骨干网络|Hmean|
|-|-|-|
|EAST^[1]^|ResNet50_vd|85.85%|
|EAST^[1]^|MobileNetV3|79.08%|
|DB^[2]^|ResNet50_vd|83.30%|
|DB^[2]^|MobileNetV3|73.00%|
LDOUBLEV's avatar
LDOUBLEV committed
40
41
42
43
44

PaddleOCR文本检测算法的训练与使用请参考[文档](./doc/detection.md)

## 文本识别算法:

LDOUBLEV's avatar
LDOUBLEV committed
45
PaddleOCR开源的文本识别算法列表:
LDOUBLEV's avatar
LDOUBLEV committed
46
- [CRNN](https://arxiv.org/abs/1507.05717)
LDOUBLEV's avatar
LDOUBLEV committed
47
48
49
50
51
- [Rosetta](https://arxiv.org/abs/1910.05085)
- [STAR-Net](http://www.bmva.org/bmvc/2016/papers/paper043/index.html)
- [RARE](https://arxiv.org/abs/1603.03915v1)
- [SRN]((https://arxiv.org/abs/2003.12294))(百度自研)

LDOUBLEV's avatar
LDOUBLEV committed
52
算法效果如下表所示,精度指标是在IIIT, SVT, IC03, IC13, IC15, SVTP, CUTE数据集上的评测结果的平均值。
LDOUBLEV's avatar
LDOUBLEV committed
53
54
55
56
57
58
59
60
61
62
63
64
65
66

|模型|骨干网络|ACC|
|-|-|-|
|Rosetta|Resnet34_vd|80.24%|
|Rosetta|MobileNetV3|78.16%|
|CRNN|Resnet34_vd|82.20%|
|CRNN|MobileNetV3|79.37%|
|STAR-Net|Resnet34_vd|83.93%|
|STAR-Net|MobileNetV3|81.56%|
|RARE|Resnet34_vd|84.90%|
|RARE|MobileNetV3|83.32%|

PaddleOCR文本识别算法的训练与使用请参考[文档](./doc/recognition.md)

LDOUBLEV's avatar
LDOUBLEV committed
67
68
## TODO
**端到端OCR算法**
LDOUBLEV's avatar
LDOUBLEV committed
69
70
71
72
73
74
75
PaddleOCR即将开源百度自研端对端OCR模型[End2End-PSL](https://arxiv.org/abs/1909.07808),敬请关注。
- End2End-PSL (comming soon)



# 参考文献
```
LDOUBLEV's avatar
LDOUBLEV committed
76
1. EAST:
LDOUBLEV's avatar
LDOUBLEV committed
77
78
79
80
81
82
83
84
@inproceedings{zhou2017east,
  title={EAST: an efficient and accurate scene text detector},
  author={Zhou, Xinyu and Yao, Cong and Wen, He and Wang, Yuzhi and Zhou, Shuchang and He, Weiran and Liang, Jiajun},
  booktitle={Proceedings of the IEEE conference on Computer Vision and Pattern Recognition},
  pages={5551--5560},
  year={2017}
}

LDOUBLEV's avatar
LDOUBLEV committed
85
2. DB:
LDOUBLEV's avatar
LDOUBLEV committed
86
87
88
89
90
91
92
@article{liao2019real,
  title={Real-time Scene Text Detection with Differentiable Binarization},
  author={Liao, Minghui and Wan, Zhaoyi and Yao, Cong and Chen, Kai and Bai, Xiang},
  journal={arXiv preprint arXiv:1911.08947},
  year={2019}
}

LDOUBLEV's avatar
LDOUBLEV committed
93
3. DTRB:
LDOUBLEV's avatar
LDOUBLEV committed
94
95
96
97
98
99
100
101
@inproceedings{baek2019wrong,
  title={What is wrong with scene text recognition model comparisons? dataset and model analysis},
  author={Baek, Jeonghun and Kim, Geewook and Lee, Junyeop and Park, Sungrae and Han, Dongyoon and Yun, Sangdoo and Oh, Seong Joon and Lee, Hwalsuk},
  booktitle={Proceedings of the IEEE International Conference on Computer Vision},
  pages={4715--4723},
  year={2019}
}

LDOUBLEV's avatar
LDOUBLEV committed
102
4. SAST:
LDOUBLEV's avatar
LDOUBLEV committed
103
104
105
106
107
108
109
110
@inproceedings{wang2019single,
  title={A Single-Shot Arbitrarily-Shaped Text Detector based on Context Attended Multi-Task Learning},
  author={Wang, Pengfei and Zhang, Chengquan and Qi, Fei and Huang, Zuming and En, Mengyi and Han, Junyu and Liu, Jingtuo and Ding, Errui and Shi, Guangming},
  booktitle={Proceedings of the 27th ACM International Conference on Multimedia},
  pages={1277--1285},
  year={2019}
}

LDOUBLEV's avatar
LDOUBLEV committed
111
5. SRN:
LDOUBLEV's avatar
LDOUBLEV committed
112
113
114
115
116
117
118
@article{yu2020towards,
  title={Towards Accurate Scene Text Recognition with Semantic Reasoning Networks},
  author={Yu, Deli and Li, Xuan and Zhang, Chengquan and Han, Junyu and Liu, Jingtuo and Ding, Errui},
  journal={arXiv preprint arXiv:2003.12294},
  year={2020}
}

LDOUBLEV's avatar
LDOUBLEV committed
119
6. end2end-psl:
LDOUBLEV's avatar
LDOUBLEV committed
120
121
122
123
124
125
126
127
@inproceedings{sun2019chinese,
  title={Chinese Street View Text: Large-scale Chinese Text Reading with Partially Supervised Learning},
  author={Sun, Yipeng and Liu, Jiaming and Liu, Wei and Han, Junyu and Ding, Errui and Liu, Jingtuo},
  booktitle={Proceedings of the IEEE International Conference on Computer Vision},
  pages={9086--9095},
  year={2019}
}
```