DeepFake Detection Projects

This repository contains projects focused on the study and detection of DeepFake images using machine learning techniques.

Dataset

DeepFake Dataset

Approaches

1. Pretrained Neural Networks & Fine-Tuning

• Utilized state-of-the-art models such as AlexNet, GoogleNet, and ViT to classify DeepFake images.
• Applied fine-tuning to improve model performance.

2. Fourier-Based SVM Classification

• Extracted frequency-domain features from DeepFake images using Fourier analysis.
• Trained Support Vector Machines (SVMs) on these features to detect AI-generated faces.
• Evaluated the effectiveness of frequency-based approaches in detecting different DeepFake generation methods.

Tools & Technologies

• Python
• PyTorch
• Jupyter Notebooks
• Scikit-learn
• Fourier Analysis techniques

Key Takeaway

These projects demonstrate practical experience in applying machine learning techniques to image forensics, combining deep learning and classical methods to detect AI-generated images.