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

请问table中的数据集STD是单纯指使用合成数据集预训练吗？还是使用STD+URD？ #8

Closed Echhoo closed 9 months ago

Echhoo commented 9 months ago

还是我应该理解成URD数据集是单纯用来做预训练的。data的：STD和ARD用来做微调，十分感谢~

TongkunGuan commented 9 months ago

还是我应该理解成URD数据集是单纯用来做预训练的。data的：STD和ARD用来做微调，十分感谢~

The pre-training experiments are conducted on URD and STD without labels, at a resolution of 32 × 128, for fair comparisons. Our text recognition network is fine-tuned at 32×128 resolution with STD or ARD.