Brian Taylor

I recently finished a Ph.D. at UCLA, where I worked with Prof. Stefano Soatto in the UCLA Vision Lab. My interests involve understanding how to discover, identify, and accurately track objects in video. For a couple of summers, I had the opportunity to dabble in text detection at Facebook.
Prior to UCLA, I obtained my B.S. from UC Berkeley, where I studied occlusion detection and put together hand-tracking experiments in Prof. Ruzena Bajcsy's Teleimmersion Lab.

contact: brianrtaylor dot mail at gmail dot com
Recent CV

Projects

Causal Video Object Segmentation from Persistence of Occlusions B. Taylor, V. Karasev, and S. Soatto. In CVPR 2015 (oral). [pdf] [project page]

We exploit occlusions to discover salient objects in video and reliably track them when occlusion cues fail.

Semantic Video Segmentation From Occlusion Relations Within a Convex Optimization Framework B. Taylor, A. Ayvaci, A. Ravichandran, and S. Soatto. In EMMCVPR 2013. [pdf] [project page]

We combine occlusions and semantic classifiers to peform pixel-level object detection and localization in video.

Teaching

CS 31: Spring 2015 Discussion Notes
CS 31: Fall 2014 Discussion Notes