Scaling Visual Search: An R&D Case Study

Scaling Visual Search: An R&D Case Study


Presented at the PHLAI in 2016.


Consumers are changing the way that they shop: the ubiquity of mobile devices has moved the customer journey from the shopping mall to the sofa. This begs the question: how can consumers discover new products if they don’t know what to search for? Curalate, a Philadelphia-based startup, is answering this question with products that leverage state-of-the-art computer vision and machine learning technology. In this talk, I discuss how Curalate’s unique research and development process produced a scalable, cutting-edge visual search technology on a startup’s time and budget. Specifically, I dive deep into our solutions to three technical challenges: building deep learning infrastructure on a budget, improving visual search accuracy using deep metric embedding, and scaling visual search with state-of-the-art compression techniques. I conclude with examples of how Curalate’s visual search technology enables brands to reach new, at-home audiences and powers novel shopping experiences for consumers.