A Constrained Latent Variable Model for Monocular Non-rigid Reconstruction
Mathieu Salzmann (NICTA)
NICTA SML SEMINARDATE: 2012-03-15
TIME: 11:15:00 - 12:00:00
LOCATION: NICTA - 7 London Circuit
CONTACT: JavaScript must be enabled to display this email address.
ABSTRACT:
In this talk, I will present my recent research on combining physical constraints with machine learning algorithms for 3D reconstruction of non-rigid surfaces from monocular images. I will focus on the formulation of a constrained latent variable model that lets us account for constraints between the different dimensions of its multivariate output. In particular, I will describe the case of isometrically deforming surfaces, and show that our constrained latent variable model provides us with an effective shape prior for monocular 3D reconstruction.
