README.md

---
title: Real vs Fake - AI Image Detector
emoji: 🔍
colorFrom: blue
colorTo: purple
sdk: gradio
sdk_version: 4.44.0
app_file: app.py
pinned: false
license: mit
---

# 🔍 Real vs Fake: AI Image Detector

Detect whether an image is real or AI-generated using perceptual features.

## How It Works

This detector uses **ARNIQA** (Attention-based distortion-aware No-Reference Image Quality Assessment)
to extract perceptual features from images. These features are then used by a trained classifier
to distinguish between real photographs and AI-generated images.

## Supported AI Models

The detector can identify images generated by:
- Stable Diffusion (various versions)
- Midjourney
- DALL-E
- And other popular generative models

## Usage

1. Upload an image (JPG, PNG)
2. The model will analyze perceptual features
3. Get instant results showing whether the image is real or AI-generated

## Model Details

- **Feature Extractor**: ARNIQA (from miccunifi/ARNIQA)
- **Classifier**: 2-layer MLP trained on GenImage dataset
- **Input Size**: Variable (automatically resized)
- **Inference**: CPU-based for universal accessibility

## Privacy

All processing happens in your browser session. Uploaded images are not stored.

## Research

Based on research using perceptual classifiers for detecting generative images.
The full research codebase is private.

## License

MIT License - Free for personal and commercial use.