Open YoussefBenCheikh opened 1 year ago
I don't think the code here works. It seems to be set up specifically for the author's local file structure and the code itself is terribly optimized and seems to have bugs. I needed to modify the data loading and preprocessing scripts to work with the Celeb-v2 dataset.
I suspect the avi extension may be located somewhere in the DFDC dataset, though I wasn't able to get access to that data.
Also, I'm not sure if the authors are planning on maintaining this codebase given that there's still an open PR since 2022.
Hi, I'm doing a project on deepfake detection, and I'm not able to reproduce your results with your pretrained model. I'm using your preprocessing script, but I noticed an avi extension in your code. I can only find CELEB-DF with mp4 extension. Can you share the testing data that gave the results in the paper. Thanks