Computer Vision

When I started my Ph. D. program, I began working on the problem of human motion tracking from images. I was particularly intrigued by the fact that humans are capable of understanding the world they see with seemingly no effort and yet computers are way far from such performance. Being able to determine what objects are present in a scene in an image or even whether a person is present in the scene are still challenging problems. And so, I became interested in model-based tracking in video sequences, probabilistic methods for approximate inference, statistical methods and machine learning techniques for detection, recognition and prediction.

During my second half of my Ph. D. I have worked on learning a deformable body model called SCAPE from a large database of laser-scans and used it to reconstruct a more realistic representation of the human body. In contrast to the simpler cylindrical body model that I was using previously, SCAPE enables the estimation of detailed body shape information in addition to the articulated pose configuration of the human subject directly from images, accounting for articulated as well as non-rigid skin deformations due to pose changes and body shape variations between individuals.

My most recent research interests lie in Computer Vision bordering Computer Graphics:

Network Reliability Analysis

During my undergraduate studies at Lafayette College, I have been working together with Lorenzo Traldi on analyzing network reliability problems using symbolic logic.