CLIP-based Deepfake Detection System

Quick Start

1. Web Interface:

   python main_demo.py --mode web

2. Run Tests:

   python main_demo.py --mode test

3. Batch Processing:

   python main_demo.py --mode batch --test-dir your_images/

Adding Test Images

Place your test images in:

• test_images/real/ - for authentic photos
• test_images/fake/ - for AI-generated images

Project Structure

• deepfake_detector.py - Core detection logic
• gradio_demo.py - Web interface
• test_utils.py - Testing and evaluation tools
• main_demo.py - Main execution script
• setup.py - This setup script

Requirements

• Python 3.8+
• PyTorch
• OpenAI CLIP
• Gradio
• Other dependencies (automatically installed)

Research Context

This is part of an MSc project on "Enhancing Deepfake Detection Robustness through Prompt Engineering and Vision-Language Models" at the University of Liverpool.

Author: Jintaek Uh Supervisor: Guangliang Cheng