Enabling Product Recommendations based on visual attributes

The proverbial statement that a picture is worth a hundred words is quite literally true! Describing a dress search with many attributes such as “full sleeve with blue dots, rounded flat neck line” etc. etc is very laborious.

The search and recommendation process especially in the fashion shopping domain is dominated by finer product attributes and user tastes.

Our deep learning model learns the finer attributes of clothing by imining visual attributes of each apparel on the site. It then understands what the user is actually looking for and recommends “visually similar” products through a deep understanding of the product attributes.