Visual Search in the Deep Learning Era
Recent breakthroughs in deep learning, including fine-grained image representation, scalable object detection, and performant on-device model inference, have enabled a new category of products to transform any image into a search query. Dmitry Kislyuk, who leads Visual Search at Pinterest, will present the latest computer vision technology which powers fast visual search across billions of images and supports hundreds of millions of users.
Dmitry Kislyuk leads Visual Search at Pinterest, where he has worked on applied computer vision and deep learning for over four years, pioneering technologies which help Pinterest understand the semantics behind every image. Most recently, he led the launch of Pinterest Lens, allowing a camera to function as a search input to find out more about anything you see, a technology which is now integrated into the core search product at Pinterest. Dmitry studied AI at U.C. Berkeley and Stanford University, and his work is focused on solving practical problems in for novel visual search systems, including visual object detection, and image embeddings for fast, fine-grained, and large-scale retrieval.