NeuralScan is a professional forensic tool designed to detect AI-generated images. It uses a Dual-Branch Deep Learning Network that analyzes both the visual RGB pixels and the frequency spectrum (FFT) of an image to determine its authenticity with high precision.
- Dual-Branch AI Engine: Combines ResNet18 (RGB) + ResNet18 (FFT Frequency Analysis) for superior detection accuracy.
- Real-Time Forensics: Instantly classifies images as "Real" or "Fake" with a confidence percentage.
- Glassmorphism UI: A modern, responsive React interface with dynamic background effects.
- Frequency Analysis: Automatically computes Fourier Transforms to spot hidden artifacts in AI-generated media.
- Framework: React (Vite)
- Styling: Tailwind CSS + Framer Motion
- Network: Axios
- API: FastAPI
- ML Engine: PyTorch & Torchvision
- Image Processing: OpenCV & Pillow (PIL)
- Node.js (v16+)
- Python (v3.9+)
- Git LFS (Required for the large model file)
Because this project contains a large AI model (~260MB), you must install Git LFS before cloning.
# Install Git LFS (Run once per computer)
git lfs install
# Clone the repo
git clone [https://github.com/devcodes28/NeuralScan.git](https://github.com/devcodes28/NeuralScan.git)
cd NeuralScan