Back To Schedule
Thursday, August 22 • 1:00pm - 1:45pm
Style-Aware Product Recommendation using Deep Ranking

Log in to save this to your schedule, view media, leave feedback and see who's attending!

Feedback form is now closed.
Interior design and home decoration heavily rely on guesswork. Customer's stylistic preference is an important factor in deciding which product to buy. Although there exist different types of styles defined by designers, labeling a scene with a style is a highly subjective task. Furthermore, customers often cannot verbalize their style preference, but can identify their preferences by looking at images. Thus, it is crucial to show products in a room context that are tailored to a customer's taste. We collect a dataset of room images labeled by interior design experts and encounter high inter-expert variability in style labels. We overcome this limitation by generating comparisons, each indicating the relative order between a pair of images. We present a deep learning based image retrieval framework to predict style from the generated comparisons. Given a seed room image, our framework predicts the style spectrum and provides a ranked list of stylistically similar room images from the catalog. Our architecture is inspired by siamese networks and extends the Bradley-Terry model to learn from comparisons. Extensive experiments show that our framework does not only accurately estimate room style, but also learns distinctive visual features reflecting style.

avatar for Esra Cansizoglu

Esra Cansizoglu

Lead Machine Learning Engineer, Wayfair
Esra is a machine learning engineer at Wayfair developing methods to provide an algorithmic understanding to room design. She holds an MS in Computer Science from Boston University and a PhD in Electrical Engineering from Northeastern University.

Thursday August 22, 2019 1:00pm - 1:45pm EDT
Room 406 595 Commonwealth Avenue, Boston, MA 02215, USA