Privacy Preserving Cloth Try-On Using Mobile Augmented Reality

Abstract

Virtual try-on applications make it possible for buyers to watch themselves wearing different garments without physically trying on them. The prevailing approach for virtual try-on has been based on virtual fitting rooms, in which several cameras are used to identify the skeleton and posture of a user in order to render a garment on the user's image. Although this approach has been implemented successfully using different techniques, the privacy of users can be compromised as some users might be reluctant to stand in front of cameras in a fitting room. This paper proposes an alternative approach that allows a customer to watch a three-dimensional (3D) model of her/him wearing garments on a personal mobile device using augmented reality (AR). Among 3D human models that are automatically generated, a model selection technique is proposed that makes it possible to find the right size model representing the anthropometric features of the user. This approach is accompanied by body customization and face generation modules to generate a realistic representation. Several quantitative experiments as well as user studies were performed to evaluate the accuracy, efficiency, usefulness, and privacy of the proposed technique.