White Mirror

Virtual Fitting Room

 

Member profile details

Organization/Company
Rice University
Team Name
White Mirror
Project Thumbnail Image
Project Title
Virtual Fitting Room
Design Challenge
Without physically trying on a garment, online clothing shoppers are unable to decide the best size or color, and therefore are likely to purchase clothes that do not fit them. This causes half of the clothes bought online to be sent back, which hurts the profit of online clothing retailers. We aim to build a software that allows online customers to visualize how a garment will look on them, as if they are in a real fitting room. When a customer shops at a website that uses our software and wants to try on a garment, the dressed body model of the customer is displayed with realistic details such as garment wrinkles. The customer can rotate the model to see how the garment fits from different views. Our software serves as a simple plug-in that adds a virtual try-on feature to existing fashion websites and can be easily integrated.
Design Summary
We provide a complete system to help online clothes shoppers visualize how a garment will look on them in 3D. Our system first uses 3D modeling to construct customized body models for online shoppers and fit whichever garments they want to try on the virtual body models.

For each garment size, we use Physically Based Simulation to simulate garment deformation on a set of training body models and apply learning to derive a mathematical deformation rule that characterizes how the garment physically deforms on different body shapes. Given any arbitrary body shape, garment simulation can be easily, instantly and accurately computed using the learned deformation rule.

In order to test our method, we created one extra-large T-shirt and one medium-sized dress. The average time for both T-shirt and dress simulation is less than 2 seconds (not including up-front run-time cost for the learning phase). Besides run time, we also defined and calculated two error measures to compare our results with PBS results quantitatively: 0.56cm and 0.99cm for T-shirt and dress respectively in terms of position and 1.37cm and 2.20cm for T-shirt and dress respectively in terms of shape. The promising experimental results show that the proposed method can achieve garment simulation in real time. The method is also highly scalable and can extend to any type of garment.

We are currently trying to extend our system to more types of garments and to improve the accuracy of garment simulation.

When a registered online customer shops at a fashion company website that uses our software and is willing to try on a garment, our software retrieves the selected garment model in the database and combines the garment model with the customer’s customized body model. In another browser window, the dressed body model of the customer is displayed with realistic details such as garment wrinkles. The customer can rotate the model to see how the garment fits from different views (e.g. front, back and side). Above all, our software serves as a simple plug-in that adds a virtual try-on feature to existing fashion websites and could be easily integrated.

-May.3rd,2014
Sponsors
IBM Flash Systems and ECE Department
Sponsor Logo
Department(s)
  • Electrical and Computer Engineering
Faculty Advisor 1 - Name
Gary Woods
Faculty Advisor 1 - Department
  • ECE
Faculty Advisor 2 - Name
Devika Subramanian
Faculty Advisor 2 - Department
  • CS
Award(s) and Recognition
IEEE-27th CASA (Computer animation and social agents) Conference Short Paper: Poster Accepted
2014 Brown School of Engineering Design Showcase and Poster Competition: Willy Revolution Award for Innovation in Engineering Design
Winner
Contact us


Oshman Engineering Design Kitchen - Rice University

6100 Main Street MS 390 | Houston, Texas | 77005

Phone: 713.348.OEDK

Email: oedk@ rice.edu


  Industry Partners