THE PRELIMINARY ORAL EXAMINATION FOR THE DEGREE OF Ph.D. IN COMPUTER SCIENCE FOR
Douglas Summers Stay
When humans comprehend information, they are able to demonstrate that comprehension by extrapolating, summarizing, comparing and translating the information. In order to build a system that can perform similar tasks, we need to develop appropriate representations. These representations are necessarily language-like, with elements corresponding to the parts of speech and grammar. I plan to develop detectors for these elements, and combine them to form a description of a scene that can be readily used to accomplish such comprehension tasks. These detectors include visual filters for scene parsing, an action grammar for activity understanding, and material segmentation for descriptive features.
In the same way that language allows us to describe a nearly unlimited number of situations, such a system is capable of the same kind of combinatorial explosion in the number of scenes it can describe, which is why I call it productive vision.
Examining Committee:
Dr. Yiannis Aloimonos - Chair
Dr. James Reggia - Dept’s Representative
Dr. Cornelia Fermuller - Committee Member
Dr. David Jacobs - Committee Member
EVERYBODY IS INVITED TO ATTEND THE PRESENTATION