This repository is dedicated to preprocessing the data from the FaceForensics++ Deepfake detection dataset. Essentially, all the videos are split into frames, the facial area of these frames are then extracted in some form, stored as image files and linked to a csv with the appropriate real/fake label. The eye-extraction/mouth-extraction/face-extraction py files extract the eye areas, the mouth areas, or the face areas respectively. By downloading the dataset, and linking it to the path variable, one can extract the appropriate facial area. Using GPU acceleration will make the process considerably faster. The requirements.txt file is slightly excessive but will work for all the repos in this project.