crnn多语种模型训练时，发现英文和数字错误率比中文等高很多，怎样改善？

Sierkinhane / CRNN_Chinese_Characters_Rec

(CRNN) Chinese Characters Recognition.

1.81k stars 537 forks source link

Open nissansz opened 5 months ago

nissansz commented 5 months ago

during multi-language model training, accuracy of English/numbers is worse than Chinese characters, how to improve?