Marcus Brubaker

Background Information

I am a Ph. D. student at the University of Toronto in the Computer Vision group.  I received by M. Sc. and Honors B. Sc. from the University of Toronto in Novermber of 2006 and June of 2004 respectively.  Though American by birth I am living in Canada by choice, by girlfriend and, more or less, by politics.

Please forgive the sparseness of this page.  More will come, time permitting.  Feel free to contact me at the obvious email address.  It's not posted here to avoid spam but it's the bit after the tilde at cs.toronto.edu.

A more formal and indepth description of my academic activites can be found in my CV.

Research Interests
My interests lie in the computational study of vision as a higher level process, independent of where and how it is implemented.  As a particular manifestation of this interest is the study of known vision systems, namely human and primate vision systems.  Related to this, I am interested in the study of non-linear dynamical systems.  Neural networks are well described as extremely non-linear dynamical systems and I believe that it is the emergent properties of these networks that give rise to consciousness and other complex mental phenomena.

My recent research projects have looked at the use of physics for tracking human motion.  We have explored abstracted models of bipedal walking in the context of monocular tracking with particle filters.  With future work we aim to generalize the approach and begin to exploit the power of physical priors to infer properties of the world.

Finally, I have a strong interest in Bayesian probabilistic methods which provide a natural and principled way to both fundamentally understand and solve complex problems.

Recent Projects
Date Title Description Online
June 2008The Kneed Walker for Human Pose TrackingA new monocular tracker using a more complex physical model including a knee and actively balanced torso.(pdf)
(avi)
June 2007Physics-based person tracking using simplified lower-body dynamicsOur CVPR paper describing a monocular, image-based tracker using abstract physics-based dynamics.(pdf)
(avi)
September 2006Physics-Based Priors for Human Pose TrackingMy Masters thesis exploring the use of abstracted physical dynamics as a prior for tracking human motion.(pdf)
May 2005 The Statistics and Biology of Higher Order Vision A neurophysiological review of higher levels of the visual processing stream in primates.  Done as a course project for CSC2541 (Topics in Machine Learning: Natural Scene Statistics) (pdf)
December 2004 MCMC for Bayesian Mixture of Gaussians without Indicators An exploration of alternative sampling techniques for Bayesian mixture models.  Done as a course project for CSC2541 (Topics in Machine Learning: Bayesian Methods). (pdf) (ps.gz)
May 2004 The Use ofBiological Vision in the Study of Computer Vision A review of approaches to applying the study of biological vision to computational vision.  Done as a course project for CS2523 (Computational Vision II). (pdf) (ps.gz)
December 2003 Graph Matching for Object Recognition A review of several graph-based object recognition methods of the time.  Done as a course project for CSC487 (Computational Vision I). (pdf) (ps.gz)

Hobbies
  • Cooking and food in general.  I love Thai and Vietnamese food.  Recently I've been trying to learn classical French sauce making.  Perhaps one day I will post some recipes.
  • Politics and economics.  I follow the news almost religiously.  I aggregate my news from several sources but favourites include:
  • Role-playing games. D&D 3rd edition, GURPS, Call of Cthulu, etc.
  • Golf.  I've played off and on all my life but have recently gotten into it more seriously.