Gianfranco Doretto / Publications

Face alignment via boosted ranking models

Wu, H., Liu, X., and Doretto, G.
Face alignment via boosted ranking models
Technical Report 2008GRC239, GE Global research, 2008. Visualization and Computer Vision Laboratory

Download

(unavailable)

Abstract

Face alignment seeks to deform a face model to match it with the features of the image of a face by optimizing an appropriate cost function. We propose a new face model that is aligned by maximizing a score function, which we learn from training data, and that we impose to be concave. We show that this problem can be reduced to learning a classifier that is able to say whether or not by switching from one alignment to a new one, the model is approaching the correct fitting. This relates to the ranking problem where a number of instances need to be ordered. For training the model, we propose to extend GentleBoost [22] to ranklearning. Extensive experimentation shows the superiority of this approach to other learning paradigms, and demonstrates that this model exceeds the alignment performance of the state-of-the-art.

BibTeX

@TechReport{wuLD08tr,
  Title                    = {Face alignment via boosted ranking models},
  Author                   = {Wu, H. and Liu, X. and Doretto, G.},
  Institution              = {GE Global research},
  Year                     = {2008},
  Address                  = {Niskayuna, NY, USA},
  Month                    = apr,
  Note                     = {Visualization and Computer Vision Laboratory},
  Number                   = {2008GRC239},
  Abstract                 = {Face alignment seeks to deform a face model to match it with the features of the image of a face by optimizing an appropriate cost function. We propose a new face model that is aligned by maximizing a score function, which we learn from training data, and that we impose to be concave. We show that this problem can be reduced to learning a classifier that is able to say whether or not by switching from one alignment to a new one, the model is approaching the correct fitting. This relates to the ranking problem where a number of instances need to be ordered. For training the model, we propose to extend GentleBoost [22] to ranklearning. Extensive experimentation shows the superiority of this approach to other learning paradigms, and demonstrates that this model exceeds the alignment performance of the state-of-the-art.},
  Bib2html_pubtype         = {Tech Reports},
  Bib2html_rescat          = {Video Surveillance, Appearance Modeling, Shape and Appearance Modeling, Integral Image Computations, Face Tracking, Face Modeling},
  File                     = {wuLD08tr.pdf:doretto\\report\\wuLD08tr.pdf:PDF;wuLD08tr.pdf:doretto\\report\\wuLD08tr.pdf:PDF},
  Owner                    = {doretto},
  Timestamp                = {2006.11.29}
}