This projects aims in detection of video deepfakes using deep learning techniques like RestNext and LSTM. We have achived deepfake detection by using transfer learning where the pretrained RestNext CNN is used to obtain a feature vector, further the LSTM layer is trained using the features. For more details follow the documentaion.
I tested the video of NT-c23 in the FF + + fake dataset, and the results were very poor. What the problem is, many of the results are correct(998_561.mp4)
PT：Your model_87_acc_20_frames_final_data.pt
I tested the video of NT-c23 in the FF + + fake dataset, and the results were very poor. What the problem is, many of the results are correct(998_561.mp4) PT：Your model_87_acc_20_frames_final_data.pt