A Deep Learning-Based System for Detecting Synthetic Images and Videos

With the increasing prevalence of deepfake technology, distinguishing real from manipulated media has become a pressing challenge.This project introduces a deep learning-based system designed to detect synthetic images and videos, helping to counteract misinformation and security threats.The system utilizes a 10-layer convolutional neural network (CNN) optimized for deepfake detection, incorporating dropout regularization to enhance model generalization.Video inputs undergo frame extraction and face detection, then classification using the trained model.The system achieves high accuracy (>99%), with strong performance in precision, recall, and an AUC-ROC score of 0.99.Developed as a full-stack application using FastAPI, TensorFlow, OpenCV, React.js, and Tailwind CSS, the solution enables efficient real-time detection and analysis.This project provides a reliable and scalable approach to identifying synthetic media, addressing the growing concerns around deepfakebased digital manipulation.