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

meijieru / crnn.pytorch

Convolutional recurrent network in pytorch

MIT License

2.4k stars 657 forks source link

during multi-language model training, accuracy of English/numbers is worse than Chinese characters, how to improve? #263

Open nissansz opened 7 months ago

nissansz commented 7 months ago

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