Generative Adversarial Network-Based Virtual Try-On with Clothing Region

Shizuma Kubo, Yusuke Iwasawa, Yutaka Matsuo

Abstract: We propose a virtual try-on method based on generative adversarial networks (GANs). By considering clothing regions, this method enables us to reflect the pattern of clothes better than Conditional Analogy GAN (CAGAN), an existing virtual try-on method based on GANs. Our method first obtains the clothing region on a person by using a human parsing model learned with a large-scale dataset. Next, using the acquired region, the clothing part is removed from a human image. A desired clothing image is added to the blank area. The network learns how to apply new clothing to the area of people’s clothing. Results demonstrate the possibility of reflecting a clothing pattern. Furthermore, an image of the clothes that the person is originally wearing becomes unnecessary during testing. In experiments, we generate images using images gathered from Zaland (a fashion E-commerce site).

Keywords: Generative Adversarial Networks, Deep Learning, Fashion, E-commerce

TL;DR: In this paper, we propose a virtual try-on method based on generative adversarial networks (GANs) by considering clothing region.