the result of the first epoch on Celeb-DF is very high

VISION-SJTU / RECCE

[CVPR2022] End-to-End Reconstruction-Classification Learning for Face Forgery Detection

MIT License

103 stars 14 forks source link

the result of the first epoch on Celeb-DF is very high #12

Closed zhongjian1999 closed 1 year ago

zhongjian1999 commented 1 year ago

Hello! Thank you for your excellent work. I use the preprogressed Celeb-DF dataset provided in #9 and start training successfully, but the result of the first epoch is: ACC 0.9712, AUC 0.9957, seems too high at the beginning. Is it normal? I noticed the program will spilt dataset into train set and test set before training, so is it necessary to split the dataset manually? Thank you for your help, and waiting for your answer.

XJay18 commented 1 year ago

Hi, I think this is normal since intra-testing results for Celeb-DF are quite saturated. The training and testing sets are split according to the List_of_testing_videos.txt file included in the dataset, so you do not need to manually split the data.