Enhancing Fashion Image Generation with GAN

Abstract—In this systematic review, we conducted a thorough examination of existing research concerning generative adversarial networks (GANs), focusing on their application in the fashion domain. Through the utilization of the Scopus database, a total of 933 publications pertaining to the use of GANs in fashion from 2017 to 2023 were identified. Ultimately, 22 relevant records were selected and examined based on the GANs' topologies, losses, inputs, and outputs. These findings formed the foundation for investigating the unique aspects of GAN methodology within the realm of fashion. Given the vast scope of fashion, a wide range of recognized methodologies have been adapted for this field. However, no approach has demonstrated clear superiority in the context of fashion. By conducting a comprehensive analysis of the literature, we identify the advancements, challenges, and research gaps in the field. Additionally, we propose potential directions for future investigations, aiming to inspire researchers to explore new horizons and drive innovation in this rapidly evolving domain. Keywords— Vanilla GAN, fashion, adversarial networks