Estimating Body Shapes from Measurements

e-Commerce now represents more than a third of apparel sales in the USA and accounts for most sales growth year on year. However, it is still hard for people to buy clothes online because they have no idea how they will look. Thus, we present an approach to model an approximation of a human body shape with a given set of body measurements to fit virtual clothes. To estimate a new body shape from body measurements, we developed two different models by using respectively linear transformations and PCA weights. Additionally, we selected the minimum number of body measurements required to estimate a similar shape as the ground truth. Finally, we evaluated our approach by comparing our results with estimations and visual evaluation via pictures and measurements taken from real people. Results show that we can approximate human shape through measurements with sufficient fidelity to simulate garment fitting.