Similarity metrics for categorization: From monolithic to category specific
Publikation: Bidrag til tidsskrift › Konferenceartikel › Forskning › fagfællebedømt
Standard
Similarity metrics for categorization : From monolithic to category specific. / Babenko, Boris; Branson, Steve; Belongie, Serge.
I: Proceedings of the IEEE International Conference on Computer Vision, 2009, s. 293-300.Publikation: Bidrag til tidsskrift › Konferenceartikel › Forskning › fagfællebedømt
Harvard
APA
Vancouver
Author
Bibtex
}
RIS
TY - GEN
T1 - Similarity metrics for categorization
T2 - 12th International Conference on Computer Vision, ICCV 2009
AU - Babenko, Boris
AU - Branson, Steve
AU - Belongie, Serge
PY - 2009
Y1 - 2009
N2 - Similarity metrics that are learned from labeled training data can be advantageous in terms of performance and/or efficiency. These learned metrics can then be used in conjunction with a nearest neighbor classifier, or can be plugged in as kernels to an SVM. For the task of categorization two scenarios have thus far been explored. The first is to train a single " monolithic" similarity metric that is then used for all examples. The other is to train a metric for each category in a 1-vs-all manner. While the former approach seems to be at a disadvantage in terms of performance, the latter is not practical for large numbers of categories. In this paper we explore the space in between these two extremes. We present an algorithm that learns a few similarity metrics, while simultaneously grouping categories together and assigning one of these metrics to each group. We present promising results and show how the learned metrics generalize to novel categories.
AB - Similarity metrics that are learned from labeled training data can be advantageous in terms of performance and/or efficiency. These learned metrics can then be used in conjunction with a nearest neighbor classifier, or can be plugged in as kernels to an SVM. For the task of categorization two scenarios have thus far been explored. The first is to train a single " monolithic" similarity metric that is then used for all examples. The other is to train a metric for each category in a 1-vs-all manner. While the former approach seems to be at a disadvantage in terms of performance, the latter is not practical for large numbers of categories. In this paper we explore the space in between these two extremes. We present an algorithm that learns a few similarity metrics, while simultaneously grouping categories together and assigning one of these metrics to each group. We present promising results and show how the learned metrics generalize to novel categories.
UR - http://www.scopus.com/inward/record.url?scp=77953185204&partnerID=8YFLogxK
U2 - 10.1109/ICCV.2009.5459264
DO - 10.1109/ICCV.2009.5459264
M3 - Conference article
AN - SCOPUS:77953185204
SP - 293
EP - 300
JO - Proceedings of the IEEE International Conference on Computer Vision
JF - Proceedings of the IEEE International Conference on Computer Vision
Y2 - 29 September 2009 through 2 October 2009
ER -
ID: 302049094