Efficient face detection by a cascaded support-vector machine expansion
PDF
We describe a fast system for the detection and localization of human faces in images using a nonlinear ‘support-vector machine‘. We approximate the decision surface in terms of a reduced set of expansion vectors and propose a cascaded evaluation which has the property that the full support-vector expansion is only evaluated on the face-like parts of the image, while the largest part of typical images is classified using a single expansion vector (a simpler and more efficient classifier). As a result, only three reduced-set vectors are used, on average, to classify an image patch. Hence, the cascaded evaluation, presented in this paper, offers a thirtyfold speed-up over an evaluation using the full set of reduced-set vectors, which is itself already thirty times faster than classification using all the support vectors.
| Author(s): | Romdhani, S. and Torr, P. and Schölkopf, B. and Blake, A. |
| Links: | |
| Journal: | Proceedings of The Royal Society of London A |
| Volume: | 460 |
| Number (issue): | 2501 |
| Pages: | 3283-3297 |
| Year: | 2004 |
| Month: | November |
| Day: | 0 |
| Series: | A |
| BibTeX Type: | Article (article) |
| DOI: | 10.1098/rspa.2004.1333 |
| Electronic Archiving: | grant_archive |
| Language: | en |
| Organization: | Max-Planck-Gesellschaft |
| School: | Biologische Kybernetik |
BibTeX
@article{2939,
title = {Efficient face detection by a cascaded support-vector machine expansion},
journal = {Proceedings of The Royal Society of London A},
abstract = {We describe a fast system for the detection and localization of human faces in images using a nonlinear ‘support-vector machine‘. We approximate the decision surface in terms of a reduced set of expansion vectors and propose a cascaded evaluation which has the property that the full support-vector expansion is only evaluated on the face-like parts of the image, while the largest part of typical images is classified using a single expansion vector (a simpler and more efficient classifier). As a result, only three reduced-set vectors are used, on average, to classify an image patch. Hence, the cascaded evaluation, presented in this paper, offers a thirtyfold speed-up over an evaluation using the full set of reduced-set vectors, which is itself already thirty times faster than classification using all the support vectors.},
volume = {460},
number = {2501},
pages = {3283-3297},
series = {A},
organization = {Max-Planck-Gesellschaft},
school = {Biologische Kybernetik},
month = nov,
year = {2004},
author = {Romdhani, S. and Torr, P. and Sch{\"o}lkopf, B. and Blake, A.},
doi = {10.1098/rspa.2004.1333},
month_numeric = {11}
}