Abstract: Deepfake technology has become a serious concern due to its potential misuse in misinformation, fraud, and privacy violations. Traditional detection methods struggle to keep up with increasingly sophisticated fake videos. This project leverages deep learning and computer vision techniques to detect DeepFake content in images and videos using the XceptionNet model. The system processes images and videos by extracting frames, preprocessing them, and passing them through a trained Xception model to classify them as real or fake. The video classification is based on majority voting of analyzed frames. The application is built using Streamlit for an interactive user interface, enabling users to upload and analyze media in real-time.Future improvements include optimizing model inference, enhancing dataset diversity, and integrating real-time DeepFake detection for live streaming applications.
Keywords: DeepFake Detection, XceptionNet, Deep Learning, Computer Vision, Image Processing, Video Processing, Frame Extraction, Streamlit, Realtime Detection, Majority Voting, Fake Video Classification, Media Analysis, Model Inference Optimization, Dataset Diversity, Live Streaming Applications.
|
DOI:
10.17148/IJARCCE.2025.14356
