Detection of deepfaked images from the FaceForensics++ dataset using a pre-trained ResNet in PyTorch. The model achieves 99% accuracy for all data splits!
You can download the model weights from here: https://drive.google.com/uc?id=1RJMdwG1Se94hUnIiUCwgGcmgoj5x3qdH&export=download