Skip to content

README.md

Latest commit

 

History

History
55 lines (37 loc) · 1.4 KB

README.md

File metadata and controls

55 lines (37 loc) · 1.4 KB

AI Image Authentication System

A forensic tool for detecting AI-generated imagery using a custom Convolutional Neural Network (CNN).

🚀 Features

  • AI Detection: Accurately classifies images as "REAL" or "AI GENERATED"
  • Forensic Reports: Generates a downloadable PDF datasheet for every analysis
  • Modern UI: Sleek, high-tech interface with orange accents and Rajdhani typography
  • Optimized: Supports CUDA acceleration and Mixed Precision training

🛠 Setup

1. Requirements

Install the necessary Python packages:

pip install torch torchvision gradio fpdf datasets tqdm Pillow

2. Model Weights

Ensure your trained model file model_epoch_10.pth is located in the project root directory.

3. Running the App

Launch the web interface:

python main.py

🧠 Architecture

The system utilizes a custom 5-Layer CNN designed specifically to identify the frequency artifacts found in synthetic media.

  • Input: 224x224 RGB
  • Features: 5 blocks (Conv2D → ReLU → BatchNorm → MaxPool)
  • Output: Sigmoid (0 = AI, 1 = Real)

📂 Project Structure

.
├── app.py                    # Gradio UI and inference logic
├── train.py                  # Training script and dataset preprocessing
├── model_epoch_10.pth        # Trained model weights
└── Forensic_Report_*.pdf     # Generated analysis outputs