Generating Artificial Images by Generative Adversary Network

Abstract

This paper presents a case study of the structure, generative adversary network (GAN). The primary goal of this project is to apply the concept of GAN (Generative Adversary Network) to generate a group of artificial images with same theme, which are expected to give a realistic view to human eyes in final stage. This independent research mainly focuses on basic structure of GAN and the improvement of its quality via implementing a variation, DCGAN. Thus, it can offer solid foundations and help to our team to focus on exploring a possibility of this fair new technology.