Skip to content

Commit 6ef2c87

Browse files
committed
[application] svtr ch model
1 parent 31ab300 commit 6ef2c87

File tree

1 file changed

+7
-2
lines changed

1 file changed

+7
-2
lines changed

applications/高精度中文识别模型.md

Lines changed: 7 additions & 2 deletions
Original file line numberDiff line numberDiff line change
@@ -9,7 +9,12 @@ PP-OCRv3是百度开源的超轻量级场景文本检测识别模型库,其中
99
- UDML:联合互学习策略;
1010
- UIM:无标注数据挖掘方案。
1111

12-
其中 *UIM:无标注数据挖掘方案* 使用了高精度的SVTR中文模型,该模型在PP-OCRv3识别的数据集上训练,精度为82.5%。本项目主要介绍SVTR中文模型如何使用。
12+
其中 *UIM:无标注数据挖掘方案* 使用了高精度的SVTR中文模型进行无标注文件的刷库,该模型在PP-OCRv3识别的数据集上训练,精度对比如下表。
13+
14+
|中文识别算法|模型|精度|
15+
| --- | --- | --- |
16+
|PP-OCRv3|SVTR_LCNet|79.4%|
17+
|SVTR|SVTR-Tiny|82.5%|
1318

1419
aistudio项目链接: [高精度中文场景文本识别模型SVTR](https://aistudio.baidu.com/aistudio/projectdetail/4263032)
1520

@@ -96,6 +101,6 @@ python3 tools/infer/predict_rec.py --image_dir="./doc/imgs_words/ch/word_1.jpg"
96101

97102
**注意**
98103

99-
- 使用PP-OCRv3识别SVTR_LCNet进行推理时,不需要使用--rec_algorithm指定算法名称,使用默认的推理方式即为PP-OCRv3识别的推理过程。但是使用SVTR算法时,需要指定--rec_algorithm='SVTR'
104+
- 使用SVTR算法时,需要指定--rec_algorithm='SVTR'
100105
- 如果使用自定义字典训练的模型,需要将--rec_char_dict_path=ppocr/utils/ppocr_keys_v1.txt修改为自定义的字典
101106
- --rec_image_shape='3, 32, 320' 该参数不能去掉

0 commit comments

Comments
 (0)