Enhancing Practicality and Efficiency of Deepfake Detection

This repository contains all the necessary code to test or train the work presented in Balafrej, I., Dahmane, M. Enhancing practicality and efficiency of deepfake detection. Sci Rep 14, 31084 (2024). https://doi.org/10.1038/s41598-024-82223-y

Preprocessing

Example for DFDC training part 0:

part_name=dfdc_train_part_0
export DFDC_DATASET_PATH=./dfdc2020/train_videos/$part_name
export DFDC_PREPROCESSED_DATASET_PATH=./dfdc_preprocessed/$part_name
python $HOME/dev/scripts/preprocessing_dfdc_videos_to_videos.py

Example for DFDC test:

echo "Processing test set"
export DFDC_DATASET_PATH=.#dfdc_test_set
export DFDC_PREPROCESSED_DATASET_PATH=./dfdc_preprocessed/test_set
python ./scripts/preprocessing_dfdc_videos_to_videos.py --is_test True

Example for CelebDF:

python scripts/preprocessing_celebdfv2_videos_to_images.py

Training

python scripts/train_network_single_image.py

Testing

python scripts/train_network_single_image.py --test