TY - JOUR
T1 - On taxonomies for multi-class image categorization
AU - Binder, Alexander
AU - Müller, Klaus Robert
AU - Kawanabe, Motoaki
N1 - Funding Information:
Acknowledgements We would like to thank Shinichi Nakajima and Ulf Brefeld for enlightening discussions. This work was supported in part by the Federal Ministry of Economics and Technology of Germany (BMWi) under the project THESEUS, grant 01MQ07018 and by DFG.
PY - 2012/9
Y1 - 2012/9
N2 - We study the problem of classifying images into a given, pre-determined taxonomy. This task can be elegantly translated into the structured learning framework. However, despite its power, structured learning has known limits in scalability due to its high memory requirements and slow training process. We propose an efficient approximation of the structured learning approach by an ensemble of local support vector machines (SVMs) that can be trained efficiently with standard techniques. A first theoretical discussion and experiments on toy-data allow to shed light onto why taxonomy-based classification can outperform taxonomy-free approaches and why an appropriately combined ensemble of local SVMs might be of high practical use. Further empirical results on subsets of Caltech256 and VOC2006 data indeed show that our local SVM formulation can effectively exploit the taxonomy structure and thus outperforms standard multi-class classification algorithms while it achieves on par results with taxonomy-based structured algorithms at a significantly decreased computing time.
AB - We study the problem of classifying images into a given, pre-determined taxonomy. This task can be elegantly translated into the structured learning framework. However, despite its power, structured learning has known limits in scalability due to its high memory requirements and slow training process. We propose an efficient approximation of the structured learning approach by an ensemble of local support vector machines (SVMs) that can be trained efficiently with standard techniques. A first theoretical discussion and experiments on toy-data allow to shed light onto why taxonomy-based classification can outperform taxonomy-free approaches and why an appropriately combined ensemble of local SVMs might be of high practical use. Further empirical results on subsets of Caltech256 and VOC2006 data indeed show that our local SVM formulation can effectively exploit the taxonomy structure and thus outperforms standard multi-class classification algorithms while it achieves on par results with taxonomy-based structured algorithms at a significantly decreased computing time.
KW - Multi-class object categorization
KW - Structure learning
KW - Support vector machine
KW - Taxonomies
UR - http://www.scopus.com/inward/record.url?scp=84863619408&partnerID=8YFLogxK
UR - http://www.scopus.com/inward/citedby.url?scp=84863619408&partnerID=8YFLogxK
U2 - 10.1007/s11263-010-0417-8
DO - 10.1007/s11263-010-0417-8
M3 - Article
AN - SCOPUS:84863619408
VL - 99
SP - 281
EP - 301
JO - International Journal of Computer Vision
JF - International Journal of Computer Vision
SN - 0920-5691
IS - 3
ER -