Data-Driven 3-D Human Body Customization With a Mobile Device

It is more convincing for users to have their own 3-D body shapes in the virtual fitting room when they shop clothes online. However, existing methods are limited for ordinary users to efficiently and conveniently access their 3-D bodies. We propose an efficient data-driven approach and develop an android application for 3-D body customization. Users stand naturally and their photos are taken from front and side views with a handy phone camera. They can wear casual clothes like a short-sleeved/long-sleeved shirt and short/long pants. First, we develop a user-friendly interface to semi-automatically segment the human body from photos. Then, the segmented human contours are scaled and translated to the ones under our virtual camera configurations. Through this way, we only need one camera to take photos of human in two views and do not need to calibrate the camera, which satisfy the convenience requirement. Finally, we learn body parameters that determine the 3-D body from dressed-human silhouettes with cascaded regressors. The regressors are trained using a database containing 3-D naked and dressed body pairs. Body parameters regression only costs 1.26 s on an android phone, which ensures the efficiency of our method. We invited 12 volunteers for tests, and the mean absolute estimation error for chest/waist/hip size is 2.89/1.93/2.22 centimeters. We additionally use 637 synthetic data to evaluate the main procedures of our approach.