search

VISION-SJTU / RECCE

[CVPR2022] End-to-End Reconstruction-Classification Learning for Face Forgery Detection
MIT License
103 stars 14 forks source link

Dataset downloading issues #16

Closed mamarasulovs closed 1 year ago

mamarasulovs commented 1 year ago

I downloaded not the full FaceForensics++ dataset, but a 9GB dataset from Kaggle, extracted faces from the videos in the "deepfakes" directory, and placed them in the training and testing subfolders. Does this method also work? Is it wrong?

XJay18 commented 1 year ago

Hi, sorry for the late reply. The dataset you download may have a different dataset structure from the original FaceForensics++. I think you could create a custom dataloader inherited from AbstractDataset.

The custom dataloader should fill self.images and self.targets properly according to your specific dataset structure. self.images is a list of strings specified the corresponding image path while self.targets is a list of integers specified the label of the corresponding image. Then, add a new key-value pair in this dictionary for registering the custom dataloader. Lastly, specified the dataloader name in the model configuration file, ie, data/name entry.

After these steps, you may use the custom dataloader for training using RECCE.

XJay18 commented 1 year ago

Close due to inactivation. Please feel free to reopen this issue if you still have related problems.