"experiments/simbricks/orchestration/simulators.py" did not exist on "145ad9ea2aef63bc11752557bc3f5e8b16b1c47d"
README.md 3.52 KB
Newer Older
sugon_cxj's avatar
sugon_cxj committed
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
# PaddleOCR
## 模型介绍
基于百度PaddleOCR的字符识别模型。
## 模型结构
整体包含检测和识别两个模型。
## 数据集
推荐使用icdar2015数据集[icdar2015](https://rrc.cvc.uab.es/?ch=4&com=downloads)

检测模型训练集文件结构
```
/PaddleOCR/train_data/icdar2015/text_localization/
  └─ icdar_c4_train_imgs/         Training data of icdar dataset
  └─ ch4_test_images/             Testing data of icdar dataset
  └─ train_icdar2015_label.txt    Training annotation of icdar dataset
  └─ test_icdar2015_label.txt     Test annotation of icdar dataset
```
识别模型训练集文件结构
```
|-train_data
  |-rec
    |- rec_gt_train.txt
    |- train
        |- word_001.png
        |- word_002.jpg
        |- word_003.jpg
        | ...
    |-ic15_data
        |- rec_gt_test.txt
        |- test
            |- word_001.jpg
            |- word_002.jpg
            |- word_003.jpg
            | ...
```
## 训练及推理
### 环境配置
[光源](https://www.sourcefind.cn/#/service-details)可拉取训练以及推理的docker镜像,在[光合开发者社区](https://cancon.hpccube.com:65024/4/main/)可下载paddle安装包。PaddleOCR推荐的镜像如下:
* 训练/推理镜像:docker pull image.sourcefind.cn:5000/dcu/admin/base/paddlepaddle:2.3.2-centos7.6-dtk-22.10.1-py37-latest

### 训练
以icdar2015数据集训练

检测模型
python3 -m paddle.distributed.launch --gpus '0,1,2,3'  tools/train.py -c configs/det/det_mv3_db.yml -o Global.pretrained_model=./pretrain_models/MobileNetV3_large_x0_5_pretrained

识别模型
sugon_cxj's avatar
sugon_cxj committed
47
python3 -m paddle.distributed.launch --gpus '0,1,2,3'  tools/train.py -c configs/rec/PP-OCRv3/en_PP-OCRv3_rec.yml -o Global.pretrained_model=./pretrain_models/en_PP-OCRv3_rec_train/best_accuracy
sugon_cxj's avatar
sugon_cxj committed
48
49
50
51
52
53

### 测试
检测模型
python3 -m paddle.distributed.launch --gpus '0,1,2,3'  tools/eval.py -c configs/det/det_mv3_db.yml -o Global.pretrained_model=./output/db_mv3/best_accuracy.pdparams

识别模型
sugon_cxj's avatar
sugon_cxj committed
54
python3 -m paddle.distributed.launch --gpus '0,1,2,3'  tools/eval.py -c configs/rec/PP-OCRv3/en_PP-OCRv3_rec.yml -o Global.pretrained_model=./output/v3_en_mobile/best_accuracy.pdparams
sugon_cxj's avatar
sugon_cxj committed
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96

### 推理
python3 tools/infer/predict_system.py --image_dir="./doc/imgs/" --det_model_dir="./ch_PP-OCRv3_det_infer/" --rec_model_dir="./ch_PP-OCRv3_rec_infer/" --use_angle_cls=false --rec_image_shape=3,48,320 --warmup=1

## 性能和准确率数据

检测模型测试
| Model | Precision | Recall |
| :------: | :------: |:------: |
| det | 0.7054 | 0.7193  |

识别模型测试
| Model | Acc | 
| :------: | :------: |
| rec | 0.6490 | 

推理
| 图片 | 检测数/识别数 | 检测时间(ms) |识别时间(s) |
| :------: | :------: |:------: |:------: |
| 00006737.jpg | 38 | 0.1446 |0.1450 |
| 00009282.jpg | 13 | 0.1412 |0.0898 |
| 00015504.jpg | 74 | 0.1891 |0.2880 |
| 00018069.jpg | 69 | 0.1790 |0.1325 |
| 00056221.jpg | 19 | 0.1326 |0.1175 |
| 00057937.jpg | 50 | 0.1504 |0.1546 |
| 00059985.jpg | 24 | 0.1372 |0.1146 |
| 00077949.jpg | 20 | 0.1347 |0.0838 |
| 00111002.jpg | 33 | 0.1446 |0.1259 |
| 00207393.jpg | 4  | 0.1296 |0.0408 |
| 1.jpg        | 2  | 0.1227 |0.0290 |
| 11.jpg       | 16 | 0.1227 |0.0817 |
| 12.jpg       | 4  | 0.1129 |0.0284 |
| french_0.jpg | 11 | 0.1191 |0.0493 |
| ger_1.jpg    | 4  | 0.1127 |0.0277 |
| ger_2.jpg    | 3  | 0.1121 |0.0278 |
| japan_1.jpg  | 3  | 0.1113 |0.0277 |
| japan_2.jpg  | 59 | 0.1551 |0.1133 |
| korean_1.jpg | 6  | 0.1161 |0.0395 |
| model_prod_flow_ch.png | 11 | 0.1182 |0.0502 |
## 参考
* [PaddleOCR](https://github.com/PaddlePaddle/PaddleOCR)