To download FaceForensics++ data, run scripts/download_ff_data.sh
(these will be videos). It is necessary to first get access to the FaceForensics++ download script from the official FaceForensics++ github repository. Once you have access to it, place it in the scripts directory like this: scripts/faceforensics_download_v4.py
To convert FaceForensics++ videos to images, run run_parallel_jobs.sh
from the /scripts
folder
To download XceptionNet, run scripts/download_xception.sh
To create the folder with base images, run fill_bases_directory
from /data/datasets.py
To run a feature collision attack on pretrained XceptionNet and FF++ data, run pipelines/xception_ff_feature_collision.py