Kamelia Aryafar and Jerry Soung – Looking Through Camera Lenses: The Application of Computer Vision at Etsy

Looking through Camera Lenses: Computer Vision and its Applications in Etsy

Online retail is a visual experience – shoppers have particular styles that they find appealing; often images are used as first order information when making shopping decisions. There are a variety of signals extracted from the images representing those items for sale by shoppers. Amongst these, color composition, image features and various quality scores serve as important cues for visual search and image ranking. Often shoppers have a palette of favorite colors, or a mental image of what they’re looking for. Moreover, images can be used to extract meaningful information for security applications such as duplicate and tamper detection.

In this talk, we present a review of the state-of-the-art mathematical models for image processing including feature and signatures extraction, matching and classification for tag recommendation, search by color and duplicate detection among other applications. We introduce the applications of these technologies at Etsy and discuss our experience and challenges integrating these technologies with our existing models.

About Kamelia Aryafar

Kamelia is a data scientist at Etsy working for the past two years with machine learning models and computer vision tools to curate a personalized experience for users. Prior to Etsy, Kamelia was a Ph.D. student at Drexel University working on machine learning techniques for information retrieval from large-scale music data sets.

About Jerry Soung

Jerry is a staff engineer with Etsy’s Risk Engineering team. He has been with Etsy for more than four years. He is the primary developer behind SCRAM — Etsy’s internal tool for detecting fraud, abuse, and policy violations in the marketplace. Recently, he’s extended this tool, working closely with the Data Science team to incorporate machine learning and other advanced techniques into the SCRAM detection repertoire.

Livestream

Can't join us in person? This event will be streamed at 8pm Eastern.

Event Details

Wednesday, May 27, 2015

7:30-8:00 pm: Registration
8:00-9:00 pm: Presentation
9:00-10:00 pm: Social Hour

Code as Craft participants are expected to abide by our Code of Conduct.