GANs-based Clothes Design

Abstract

Machine learning have been recently applied to multiple areas, including fashion. Fashion design by generated images makes it possible to inherit design without fashion designer and get inspiration, however, little research has been done on usage of machine learning for creation of designer clothing. The state-of-the-art works aim for high-definition output images. However in fashion design image generation, it has not been thoroughly investigated to what extent the quality of the generated image should be provided to the pattern makers that draw the costume pattern from the design images. Therefore, in this paper we propose a method of generation of clothing images for pattern makers using Progressive Growing of GANs (P-GANs) and conduct a user study to investigate whether the different image quality factors such as epoch and resolution affect the participants' confidence score. We discuss the results and possible applications of the developed method.