TRACKING WITH GPDM

Reinier Noorda

Department of Information and Computing Sciences, Utrecht University

Abstract

This experimentation project is about the analysis of GPLVM (Gaussian Process Latent Variable Model) and GPDM (Gaussian Process Dynamic Model). A few trackig methods make use of GPDM, which is an extension to GPLVM. Given a training set with high dimensional data, GPDM can give us a lower-dimensional representation of the set, which later can function as the prior term in the tracking objective function. In our experiments we try to find out to what extent the lower-dimensional representation of human poses are unique and to what extent this influences the results of the method in real-time tracking applications, along with other strengths and weaknesses.

Experimentation Project Report

Tracking with GPDM [PDF]
Reinier Noorda

System Pipeline

Results:

References

• Urtasun, R., Fleet, D.J. & Fua, P. (2006). 3D People Tracking with Gaussian Process Dynamical Models. CVPR, 238-245.
• Jack M. Wang, David J. Fleet, Aaron Hertzmann, Gaussian Process Dynamical Models for Human Motion," IEEE Transactions on Pattern Analysis and Machine Intelligence, vol. 30, no. 2, pp. 283-298, Feb. 2008, doi:10.1109/TPAMI.2007.116
• Neil Lawrence (January 2006). Gaussian Process Latent Variable Model