Image-to-Image Attire Transfer for Virtual Trial Room

Virtual trial room is a lucrative tool for online-based attire shopping. Developing such a system is very challenging as it requires robustness from a user's point of view. In this paper, we present a technique for image-to-image attire transfer using Generative Adversarial Networks (GAN) and image processing methods that can transfer the clothing from a person's image to another person's image. Our system takes two images: (1) a full-length image of the user, and (2) an image of another person with a target clothing. Our method then produces a new image of the user with the targeted outfit while keeping the shape, pose, action, and identity of the user unchanged. For this purpose, we exploited the Liquid Warping GAN for domain transfer and U-Net with Grab-cut for segmentation. We illustrate the results of our work in this paper and the outcomes show that our approach performs very satisfactorily for image-to-image attire transfer. We believe this work will contribute to the clothing sector of our e-commerce system by making the shopping smoother, as it will be able to transfer a human model's outfit to a buyer's body in the image and thus helping him/her in deciding a suitable product to purchase.