# RDCReconstruction

**Repository Path**: kelikeli/RDCReconstruction

## Basic Information

- **Project Name**: RDCReconstruction
- **Description**: No description available
- **Primary Language**: Unknown
- **License**: Not specified
- **Default Branch**: main
- **Homepage**: None
- **GVP Project**: No

## Statistics

- **Stars**: 0
- **Forks**: 0
- **Created**: 2024-10-20
- **Last Updated**: 2024-10-23

## Categories & Tags

**Categories**: Uncategorized

**Tags**: None

## README

# Enhancing DeepFake Robustness via Real Discrete Codebook Reconstruction

## Environment
This code is implemented in PyTorch, and we have tested the code under the following environment settings:
- python = 3.7.9
- torch = 1.12.0
- torchvision = 0.13.0

## Dataset
[FaceForensics++](https://github.com/ondyari/FaceForensics)

## Usage Steps
### 1. Train Deepfake Detector
Train detectors for generating datasets with occlusion and adversarial samples.

**Train deepfake detector**
- detection/train.py

**Test deepfake detector**
- detection/test.py

**Constructing an occlusion dataset using cam**
- detection/mask_image_process.py

**Generating Adversarial Samples**
- detection/adversarial_examples_process.py

### 2.Train VQGAN
**stage1:Constructing Real Discrete Codebook (RDC) using real face images.**
- RDC_Reconstruction/training_vqgan.py

**stage2:Reconstruction of forged face images with occlusion using RDC.**
- RDC_Reconstruction/training_transformer.py