A Virtual Wardrobe: A Computer Vision Approach Towards Online Shopping

Abstract

By 2020, there will be over two billion online shoppers worldwide. Led by the massive investments of online retailers, such as Amazon and Walmart, online shopping is redefining the very essence of shopping. However, this has resulted in billions of dollars being wasted annually over handling the shipping and returns of undesired clothing. In this project, we explore several techniques and algorithms from a myriad of fields, such as computer vision, image segmentation, and mixed reality, in order to create a unique virtual wardrobe. This wardrobe allows users to “try-on” multiple clothing items in the comfort of their homes, before purchasing them online. In the following report, we will first overview existing technologies and advancements in the fields of computer vision and image processing to provide context to our project, as well as provide a foundation for its hardware and software development. We will then highlight our final design which consists of a green screen segmentation driven algorithm, coupled with a color wheel mobile app and an external monitor setup. We will also discuss the various design choices and optimizations made throughout the process of bringing a fitting room to the comfort of one’s home. We hope that our work will pave the way for more computer-vision/mixed-reality based approaches to wardrobe fitting in the future.