Deep Learning in Face Synthesis: A Survey on Deepfakes

Abstract

Deepfake stemming from the combined words of "deep learning" and "fake", refers to a type of fake images and video generation technology based on artificial intelligence. In recent years, with the continuous development of deep learning, especially auto-encoders and generative adversarial networks, deepfake has made tremendous progress, resulting in the emergence of some easy-to-use application software in the market, reducing the application difficulty of face-synthesis technology. This paper provides a thorough review of deepfake, introducing its development in the past three years.