Deepfake Detection Using Binary Neural Networks (BNNs)

This project implements a deepfake detection model using the BNext architecture on the CIFAKE dataset. The goal is to classify images as real or AI-generated synthetic images.

Setup

Step 1: Clone the repository

Run this to clone the repository

git clone https://github.com/MYAzrak/CV-Project-G5.git

Step 2: Install Dependencies

Install the required dependencies by running:

pip install -r requirements.txt

Step 3: Download the CIFAKE dataset

Download the CIFAKE dataset by cloning their repository.

git clone https://github.com/jordan-bird/CIFAKE-Real-and-AI-Generated-Synthetic-Images

Step 4: Download best_model.pth

Download the best pre-trained model to use it in testing from 'Pre-trained model' release.

Step 5: Modify the paths

Modify train_dir, test_dir, and model_path in config.py to point to your specific dataset and pre-trained_model paths.

Step 6: Try the model

Run train.py, test.py, or inference.py.

Collaborators

This project was worked on with the contributions of arcarum, Adithya, and Mohammad Yasir Azrak.