FaceForensics++ Enhanced Toolkit

Overview

This repo is an enhanced toolkit with some updated methods processing original FaceForensics++ dataset.

The FaceForensics++ dataset is designed for facial manipulation detection, offering real and altered videos using methods like FaceSwap and DeepFakes. It’s widely used to benchmark deepfake detection models.

BUT~

There exists some awkward problems when you first dive into the field ofDeepfake Detection, especially Deepfake binary mask location, like me🫠🫠

• How to download original dataset completely and uninterruptedly.
• How to extract frames and masks form videos which can be interrupted and finally no-repeated.
• How to convert extracted masks to binary masks

Note

These functions maybe unobtrusive, but they can greatly increase your efficiency and improve your workflow progress!

Quick Start

There are 4 files totally.

Tip

1. To run them on your own machine, you will need to modify args_config.
2. Just run it!

1. download dataset script. download-FaceForensics.py
2. extract frames from videos (no repeat). extract_frame.py
3. extract binary masks from mask_videos (no repeat). extract_frame.py
4. convert a mask to binary mask. generate_binary_mask.py

Acknowledgement

Many thanks to the FaceForensics++ dateset for contributing to the field of Deepfake Detection!

You can get original methods on their official repository FaceForensics.