The Taub Faculty of Computer Science Events and Talks

Cameras are ubiquitous everywhere and hold the promise of significantly changing the way we live and interact with our environment. Dynamic vision systems are uniquely positioned to address the needs of a growing segment of the population. Smart environments that are aware of user activities would enable an aging population to carry on independent lives for as long as possible. The power of geometric invariants to provide solutions to computer vision problems towards realizing the above potential has been recognized for a long time. On the other hand, dynamics-based invariants remain largely untapped. Yet, visual data come in streams: videos are temporal sequences of frames, images are ordered sequences of rows of pixels, and contours are chained sequences of edges. In this talk, I will show how making this ordering explicit allows to exploit dynamics-based invariants to efficiently capture useful information from video and image data. In particular, I will show how to use these invariants to perform data association, segmentation, and classification in the context of computer vision applications including multi-camera tracking, crossview activity recognition, and video segmentation.