This repository is directly taken from Fast adversarial training using FGSM by Wong, et. al. I adapted the code to be used for the FaceForensics++ Dataset
Taken from the original repository:
"The framework used is a modified version of the Free Adversarial Training repository, which in turn was adapted from the official PyTorch repository."
Install requirements with pip install -r requirements.txt
Taken from original repository (underline denotes my change):
Scripts to robustly train an FaceForensics++ classifier for epsilon radii of 2/255 and 4/255 are provided in
run_fast_2px.sh
andrun_fast_4px.sh
. These run the main code modulemain_free.py
using the configurations provided in theconfigs/
folder. To run the 50 step PGD adversary with 10 restarts, we also providerun_eval.sh
. All parameters can be modified by adjusting the configuration files in theconfigs/
folder.
I provide the job submission script I used (adv_train.pbs) to start phase 1 of training.