This repo trained Chinese text recognition model. The model is designed for Chinese text recognition of single line and is suitable for scene, web and document data. It may not works well with handwriting data. For multiple lines text image, you need to use a text detector.
ACC/NED | |
---|---|
Scene | 0.6676/0.8094 |
Web | 0.6567/0.8031 |
Document | 0.9700/0.9923 |
If you want to use the model and code, please refer to https://github.com/yaoxiaoyuan/mimix.
@misc{mimix,
title={mimix},
author={Xiaoyuan Yao},
year={2021}
}