metadata
title: Deepfake Detection Model
emoji: 🛡️
colorFrom: blue
colorTo: indigo
sdk: docker
app_file: backend/app.py
app_port: 7860
pinned: false
DeepGuard: AI-Powered Deepfake Detection
DeepGuard is a state-of-the-art, privacy-focused tool designed to detect AI-generated images with 96.97% accuracy. It runs entirely on your local machine using a Hybrid Multi-Branch Neural Network.
🚀 Quick Links
- 📝 Overview & How it Works
- ⚡ Getting Started Guide
- 🏗️ System Architecture
- 🔒 Security & Privacy
- 🛠️ Backend API
- 🎨 Frontend Guide
🏆 Current Performance (Mark-V)
| Metric | Score | Note |
|---|---|---|
| Accuracy | 96.97% | Tested on Universal Dataset |
| Reliability | Generative | Wide coverage of generation methods |
| FPS | ~25 | Real-time analysis on GPU |
📦 Features
- Multi-Branch Detection: Combines RGB, Frequency (FFT), Patch analysis, and Vision Transformers.
- Defense-in-Depth: Automatically detects C2PA credentials and invisible watermarks (Stable Diffusion).
- Local-First: No data ever leaves your computer.
- History Tracking: Keep a local log of your scans.
💻 Quick Install
git clone https://github.com/your-username/DeepGuard.git
cd DeepGuard/backend
python -m venv venv
source venv/bin/activate
pip install -r requirements_web.txt
python app.py
Open http://localhost:7860 in your browser.
For full documentation, please visit the Documentation Folder.