/AVT2-DWF

AVT2-DWF: Improving Deepfake Detection with Audio-Visual Fusion and Dynamic Weighting Strategies

Primary LanguagePython

AVT2-DWF: Improving Deepfake Detection with Audio-Visual Fusion and Dynamic Weighting Strategies

This repository contains the implementation of AVT2-DWF method proposed in the paper

Links: [PDF]

Network Architecture

Dependencies

  1. torch

Prepare data

  1. Data directory

    /data/real/{videoname}.mp4  
/daata/fake/{videoname}.mp4

  2. Once the videos have been placed at the above mentioned paths, run python pre-process.py --out_dir train and python pre-process.py --out_dir test for pre-processing the videos.

  3. After the above step, you can delete pyavi, pywork, pyframes and pycrop directories under train and test folders. (Do not delete pytmp folder please!)

  4. Collect video paths in csv files by running python write_csv.py --out_dir . command.

Thanks to the code available at https://github.com/abhinavdhall/deepfake/tree/main/ACM_MM_2020

Training

python all_process.py --out_dir "train dataset path" --gpu 1 --resume false

Testing

python all_process.py --test /checkpoints --out_dir "test dataset path"