Image Synthesis and Editing with Generative Adversarial Networks (GANs): A Review

Abstract

Recently, as many deep learning models are emerging, deep learning has achieved great success in the field of artificial intelligence(AI). Especially, the Generative adversarial networks (GANs) based on zero-sum game theory has become a new research hot spot in the field of deep learning. The significance of the GAN model is that it can generate realistic data through unsupervised learning. Based on the conceptual and theoretical framework of the generative adversarial network, GANs models and their application result in tremendous success among different areas, especially in image synthesis and editing. This paper visualizes the data structures of various kinds of GANs models in 3D and discusses the variational GAN models with respect to their improvements in the applications. As the GANs have superior learning ability, strong plasticity, great potential for improvement, and a wide application range, this paper prospects the possible applications of the GANs in the near future.