Stochastic Meta-Descent for Tracking Articulated Structures

M. Bray, E. Koller-Meier, N. N. Schraudolph, and L. Van Gool. Stochastic Meta-Descent for Tracking Articulated Structures. In IEEE Workshop on Articulated and Nonrigid Motion, Conference on Computer Vision and Pattern Recognition (CVPR), Washington, D.C., 2004.
Latest version

Download

3.3MB	737.8kB	3.9MB

Abstract

Recently, an optimization approach for fast visual tracking of articulated structures based on Stochastic Meta-Descent (SMD) has been presented. SMD is a gradient descent with local step size adaptation that combines rapid convergence with excellent scalability. Stochastic sampling helps to avoid local minima in the optimization process. We have extended the SMD algor ithm with new features for fast and accurate tracking by adapting the different step sizes between as well as within video frames and by introducing a robust likelihood function which incorporates both depths and surface orientati ons. A realistic deformable hand model reinforces the accuracy of our tracker. The advantages of the resulting tracker over state-of-the-art methods are corroborated through experiments.

BibTeX Entry

@inproceedings{BraKolSchVan04,
     author = {Matthieu Bray and Esther Koller-Meier and
               Nicol N. Schraudolph and Luc Van~Gool},
      title = {\href{http://nic.schraudolph.org/pubs/BraKolSchVan04.pdf}{
               Stochastic Meta-Descent for Tracking Articulated Structures}},
  booktitle = {IEEE Workshop on Articulated and Nonrigid Motion,
               Conference on Computer Vision and Pattern Recognition (CVPR)},
    address = {Washington, D.C.},
       year =  2004,
   b2h_type = {Other},
  b2h_topic = {>Stochastic Meta-Descent, Computer Vision},
   b2h_note = {<a href="b2hd-BraKolSchVan07.html">Latest version</a>},
   abstract = {
    Recently, an optimization approach for fast visual tracking of
    articulated structures based on Stochastic Meta-Descent (SMD)
    has been presented.  SMD is a gradient descent with local step size
    adaptation that combines rapid convergence with excellent scalability.
    Stochastic sampling helps to avoid local minima in the optimization
    process. We have extended the SMD algor ithm with new features for
    fast and accurate tracking by adapting the different step sizes
    between as well as within video frames and by introducing a robust
    likelihood function which incorporates both depths and surface
    orientati ons. A realistic deformable hand model reinforces the
    accuracy of our tracker. The advantages of the resulting tracker
    over state-of-the-art methods are corroborated through experiments.
}}