未验证 提交 f3d33aaa 编写于 作者: M MissPenguin 提交者: GitHub

Update README_en.md

上级 c0418ad8
......@@ -123,7 +123,7 @@ For use of [LSVT](https://github.com/PaddlePaddle/PaddleOCR/blob/develop/doc/doc
* Note: For the training and evaluation of the above DB model, post-processing parameters box_thresh=0.6 and unclip_ratio=1.5 need to be set. If using different datasets and different models for training, these two parameters can be adjusted for better result.
For the training guide and use of PaddleOCR text detection algorithms, please refer to the document [Text detection model training/evaluation/prediction](./doc/doc_en/detection_en.)
For the training guide and use of PaddleOCR text detection algorithms, please refer to the document [Text detection model training/evaluation/prediction](./doc/doc_en/detection_en.md)
## Text recognition algorithm
......
Markdown is supported
0% .
You are about to add 0 people to the discussion. Proceed with caution.
先完成此消息的编辑!
想要评论请 注册