Image Understanding Research at Rochester

Chrisopher M. Brown and Randal C. Nelson
Department of Computer Science
University of Rochester
Rochester, NY 14627

Abstract: Animate (real time, purposeful, active) vision requires parallel computing of several varieties (MIMD, SIMD, pipelined). This year's research has been devoted, on the applications side, to sophisticated real time vision algorithms in the areas of gaze control, optic flow analysis, colored object recognition, and location, and selective attention. This year the parallel computing environment has ceen able to support users on the Psyche and Platinum NUMA operating systems, the Zebra and Zed pipeline parallel programming tools, and the Instant-replay and Moviola debugging and performance monitoring toolkits. Goals for next year include integrating cognitive planning with dextrous manipulation and real time vision, and incorporating increasingly sophisticated multi-agent interaction in animate vision systems.