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

eragonruan / text-detection-ctpn

text detection mainly based on ctpn model in tensorflow, id card detect, connectionist text proposal network

MIT License

3.43k stars 1.34k forks source link

during multi-language model training, accuracy of English/numbers is worse than Chinese characters, how to improve? #498

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?