This repository provides code for detecting deepfake images and videos using pre-trained models.
Run pip install -r requirements.txt to install all the required packages Download the model weights by running the download_weights.py script Make sure all codes and model weights are in same directory
Run the webui code depending on your requirement (Image or Video), upload the required media, and click on Submit Prediction will be generated accordingly.
EfficientNet was used for the base feature extraction. A small portion of the FaceForensics++ dataset was used. Only around 8000 images were used for detection. Accuracy can be improved using more images and more variety.