AI Deepfake Detection System

DeepFake detector is an AI-based deepfake detection system designed to identify manipulated facial videos using deep learning. The system extracts representative video frames, preprocesses facial regions, and classifies videos as real or fake using a fine-tuned ConvNeXt-Tiny model. To improve transparency and trust, Grad-CAM visualizations highlight the regions that influenced the model's prediction, enabling users to understand why a video was classified as manipulated. The project focuses entirely on visual analysis without relying on audio or lip-sync features.