Key body landmarks from the photos are processed to create a 3D model. The 3D model allows us to capture key body shape characteristics when producing a size recommendation and allows us to suggest a larger or smaller size, depending on the product’s fit intent. The 3D model generation pipeline is so robust that it will even work for shoppers with older smartphone cameras.

We use various data sets to produce accurate size and fit recommendations. We start by collecting brand-specific fit standard data, including core body measurements (fit form, avatar, live models) and grading rules. Then we analyze season or collection-specific tech packs to understand the fit intent of each style. Our size recommendation algorithms map this data to a customer’s body data and fit preference to produce the optimal recommended size.

Our algorithms generate a geometrically correct virtual try-on that alters the shape of the body. In addition to 3D body mesh enhancements and body tracking, a new cloth simulation machine learning algorithm recreates the texture of fabrics. The semantic generation module modifies the person’s segmentation map to identify the area on the body that should be covered with the target clothes and warps the clothing mask accordingly.

Our technology utilizes sophisticated algorithms to analyze individual body measurements and shapes, offering personalized sizing recommendations. By accurately determining the right fit for each unique body type, AI body shape detection significantly mitigates the issue of ill-fitting garments, a common cause of returns in online shopping. Consequently, this advancement not only enhances customer satisfaction by ensuring a better fit but also reduces the frequency of returns, benefiting both consumers and retailers in terms of convenience and cost-efficiency.