Preconditioning in Computer Vision and Robotics

2011-12
Preconditioning in Computer Vision and Robotics
Frank Dellaert, School of Interactive Computing
January 17, 2012

I will talk about tracking many features in many images, and why the extracted tracks generally lie on low-dimensional linear subspaces, or near them. The problem is that some of the tracks will be incorrect, so now the problem becomes one of finding the low-rank matrices.

Preconditioning in Computer Vision and Robotics: Support Graph Theory has practical implications for computer vision. I will talk about recent work we did before we even knew about support graph theory, and perhaps this will lead to interesting collaborations.