Pattern Pattern
Client Case Study: Virtual Apparels Try-On
Client Case Study Virtual Apparels Try On

Client Case Study: Virtual Apparels Try-On

Industry tag: ECOMMERCE

Note: We take customer confidentiality very seriously. Our customer kindly asked us not to be named in this case study, but approved the content below.

Client Case Study Overview

The cost of apparel returns amounted to $101 billion, according to the National Retail Federation. Retailers have been trying to perfect virtual try-on for some years.


Our client’s solution aims to allow customers to get perfect fitting clothes, by virtually trying on clothes, with just a picture of themselves. The application scans the customer’s body and determines the right clothing size.


The application needs a substantial amount of multiple body types scan data, to accurately predict sizes from pictures. A too limited set of data points may strongly be biased, not sufficiently granular and result in inaccurate size estimations.

SOLUTION: Large-scale training data for AI virtual fitting model

The goal is to make different body sizes recognizable to machines and map clothes onto bodies. With the huge diversity of existing body shapes, the starting point is data. Our annotators trace with key points multiple body shape images. Then, they validate the 3D shape produced by the model (mesomorph, ectomorph, endomorph…) to determine size ranges and fit intent.


77% Reduction in refund or replacement due to wrong fit issue

62% Improvement in conversion rate compared to the traditional size chart

94% Global conversion rate expected after implementing the AI feature


At SmartOne, we’re passionate about the potential of AI and machine learning. We firmly believe in the power of quality datasets to drive innovation and transformative solutions in this space. Our dedicated team offers an array of services designed to assist AI teams in refining and customizing their datasets.

Request a free cost estimate for your next project or check out our data annotation services.