BYOL Yaklasimini Kullanarak MTCNN ile Yakalanan Yüzlerde Deepfake Tespiti Deepfake Detection on Faces Captured with MTCNN, by Using the BYOL Approach

Emerging artificial intelligence technologies bring new opportunities to our lives, but also introduce new challenges such as Deepfake technology. Deepfake is an Artificial Intelligence application that mimics a person's appearance, expressions, and voice to make it appear as if they are performing actions or saying things they haven't in reality. Various Artificial Intelligence approaches, including the Resnet50-enhanced "Bootstrap Your Own Latent" (BYOL) model, have been developed for detecting Deepfakes. BYOL is a self-supervised learning approach that does not rely on labeled data. Two neural networks, an online network and a target network, interact and learn from each other to enhance representation learning. BYOL and Resnet extract meaningful features like facial expressions, eyes, and nose for Deepfake detection, which are then classified against real images using a Binary Classifier. Additionally, Multi-task Cascaded Convolutional Neural Networks" (MTCNN) facilitate face detection within images to streamline the process. These methods work in coordination to detect Deepfake images effectively.