Study on Various GAN models and Datasets for Deepfake Detection

초록

Because generative adversarial networks are better at creating high-quality images and videos, deepfake images and videos have become a critical problem today. Many tech giants have tried to develop algorithms to spot deepfakes to solve this problem; however, they have focused on different artifacts and GAN models. They have also examined various experimental scenarios. It isn't easy to compare algorithms in general environments because there are many different ways to test them. This study explains how deepfake detection datasets can help in this endeavor. First, we provide a brief overview of deepfake detection datasets with traditional GAN models. In addition to the conventional deepfake detection model, we explain how to validate the deepfake detection algorithm for recent GAN models using the acquisition method.

키워드