Physics-Based People Tracking
Research Overview
Kinematic models have been used extensively to help constrain
monocular 3D tracking of human pose. While many such approaches
have been successful they exhibit common problems, many of which are
a consequence of not ensuring that the kinematic models obey fundamental
physical constraints.
We have begun to explore a different class of human motion models
for visual tracking that are parameterized using dynamics.
Related Publications
-
The Kneed Walker for human pose tracking.
(pdf)
© IEEE
Brubaker, M.A. and Fleet, D.J.
IEEE Conference on Computer Vision and Pattern Recognition (CVPR),
Anchorage, 2008.
-
Physics-based person tracking using simplified lower-body dynamics.
(pdf)
© IEEE
Brubaker, M., Fleet, D.J., and Hertzmann, A.
IEEE Conference on Computer Vision and Pattern Recognition (CVPR),
Minneapolis, 2007.
Return to David Fleet's home page.