Deepfakes Video classification via CNN, LSTM, C3D and triplets [IWBF'20]

Built a deep learning-based system to detect deepfakes in images and videos using CNNs and facial forensics. Leveraged models like XceptionNet for feature extraction and classification. Trained on datasets like Face Forensics++ for accurate real vs. fake detection.

Developed a deep learning-based system to detect AI-generated deepfake videos using ResNeXt CNN for feature extraction and LSTM-based RNN for video classification, capable of identifying both replacement and re-enactment deepfakes.

🎥 CyberEye is an AI-powered Flask web app for detecting deepfakes in videos by analyzing facial landmarks and consistency. Upload videos, review frame-by-frame analysis, and get visual classification results with an intuitive interface.

This project uses a deep learning pipeline to detect deepfakes in images and videos. The core is a CNN classifier trained on real and synthetic data to detect subtle visual artifacts.for videos, we use OpenCV to extract frames, apply the trained CNN to each frame, and aggregate the scores for a final classification.