DeepGuard — AI & Deepfake Image Detection System

A Progressive Web App for real-time detection of AI-generated and deepfake images using a custom fine-tuned Swin-Tiny model.

Architecture

• Frontend: React + Vite (PWA)
• Backend: Node.js + Express (API Gateway)
• ML Service: Python + Flask (Inference)

Features

• Custom fine-tuned Swin-Tiny deepfake detection model
• Grad-CAM heatmap explainability
• Error Level Analysis (ELA) forensic visualization
• Camera capture support
• Dynamic analysis explanation

Setup

ML Service

cd ml-service
python -m venv venv
source venv/Scripts/activate  # Windows Git Bash
pip install -r requirements.txt
python app.py

Node Backend

cd node-backend
npm install
npm run dev

React Frontend

cd react-frontend
npm install
npm run dev

Model

Download the fine-tuned model checkpoint from: After downloading, create a model folder and paste it in the ml-service folder https://drive.google.com/file/d/1YnQ3ZGG_TFQVB0o3jTrE5I6Zu6cDy4-l/view?usp=drive_link

Place it at: ml-service/models/best_swin_tiny_full.pth