search

TongkunGuan / CCD

[ICCV2023] Self-supervised Character-to-Character Distillation for Text Recognition
https://openaccess.thecvf.com/content/ICCV2023/papers/Guan_Self-Supervised_Character-to-Character_Distillation_for_Text_Recognition_ICCV_2023_paper.pdf
140 stars 7 forks source link

我想请问您在进行finetune的时候，准确率是从0开始提升呢还是从一定程度(远大于0)开始提升 #14

Open Echhoo opened 5 months ago

TongkunGuan commented 5 months ago

No description provided. pretraining: pretext task (no recognition) funtuning: text recognition(initial accuracy=0)