Event Location: GHC 2109

Abstract: In this thesis, I aim to present a new perspective on looking at classical problems
in Computer Vision. Instead of spending efforts on finding novel objective
functions and improving existing optimization algorithms, I aim to demonstrate that
formulating those problems as policy learning yields more flexible and efficient algorithms.
Our recently published work, Supervised Descent Method (SDM), is an
example thereof. SDM has been successfully applied to several problems, such as,
rigid tracking, non-rigid tracking, pose estimation, and has demonstrated superior
performance to traditional approaches. Especially, in facial feature tracking SDM
has achieved state-of-the-art results. Our ongoing research focuses on several extensions
of SDM beyond computer vision.

Committee:Fernando De la Torre, Chair

Kris Kitani

Srinivasa Narasimhan

Aleix Martinez, Ohio State