Spatial pyramids and two-layer stacking SVM classifiers for image categorization: A comparative study

Azizi Abdullah, Remco C. Veltkamp, Marco A. Wiering

Research output: Chapter in Book/Report/Conference proceedingConference contribution

19 Citations (Scopus)

Abstract

Recent research in image recognition has shown that combining multiple descriptors is a very useful way to improve classification performance. Furthermore, the use of spatial pyramids that compute descriptors at multiple spatial resolution levels generally increases the discriminative power of the descriptors. In this paper we focus on combination methods that combine multiple descriptors at multiple spatial resolution levels. A possible problem of the naive solution to create one large input vector for a machine learning classifier such as a support vector machine, is that the input vector becomes of very large dimensionality, which can increase problems of overfitting and hinder generalization performance. Therefore we propose the use of stacking support vector machines where at the first layer each support vector machine receives the input constructed by each single descriptor and is trained to compute the right output class. A second layer support vector machine is then used to combine the class probabilities of all trained first layer support vector models to learn the right output class given these reduced input vectors. We have performed experiments on 20 classes from the Caltech object database with 10 different single descriptors at 3 different resolutions. The results show that our 2-layer stacking approach outperforms the naive approach that combines all descriptors directly in a very large single input vector.

Original languageEnglish
Title of host publicationProceedings of the International Joint Conference on Neural Networks
Pages5-12
Number of pages8
DOIs
Publication statusPublished - 2009
Externally publishedYes
Event2009 International Joint Conference on Neural Networks, IJCNN 2009 - Atlanta, GA
Duration: 14 Jun 200919 Jun 2009

Other

Other2009 International Joint Conference on Neural Networks, IJCNN 2009
CityAtlanta, GA
Period14/6/0919/6/09

Fingerprint

Classifiers
Support vector machines
Image recognition
Learning systems
Experiments

ASJC Scopus subject areas

  • Software
  • Artificial Intelligence

Cite this

Abdullah, A., Veltkamp, R. C., & Wiering, M. A. (2009). Spatial pyramids and two-layer stacking SVM classifiers for image categorization: A comparative study. In Proceedings of the International Joint Conference on Neural Networks (pp. 5-12). [5178743] https://doi.org/10.1109/IJCNN.2009.5178743

Spatial pyramids and two-layer stacking SVM classifiers for image categorization : A comparative study. / Abdullah, Azizi; Veltkamp, Remco C.; Wiering, Marco A.

Proceedings of the International Joint Conference on Neural Networks. 2009. p. 5-12 5178743.

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abdullah, A, Veltkamp, RC & Wiering, MA 2009, Spatial pyramids and two-layer stacking SVM classifiers for image categorization: A comparative study. in Proceedings of the International Joint Conference on Neural Networks., 5178743, pp. 5-12, 2009 International Joint Conference on Neural Networks, IJCNN 2009, Atlanta, GA, 14/6/09. https://doi.org/10.1109/IJCNN.2009.5178743
Abdullah A, Veltkamp RC, Wiering MA. Spatial pyramids and two-layer stacking SVM classifiers for image categorization: A comparative study. In Proceedings of the International Joint Conference on Neural Networks. 2009. p. 5-12. 5178743 https://doi.org/10.1109/IJCNN.2009.5178743
Abdullah, Azizi ; Veltkamp, Remco C. ; Wiering, Marco A. / Spatial pyramids and two-layer stacking SVM classifiers for image categorization : A comparative study. Proceedings of the International Joint Conference on Neural Networks. 2009. pp. 5-12
@inproceedings{01664626be654b6485125f028200bb2a,
title = "Spatial pyramids and two-layer stacking SVM classifiers for image categorization: A comparative study",
abstract = "Recent research in image recognition has shown that combining multiple descriptors is a very useful way to improve classification performance. Furthermore, the use of spatial pyramids that compute descriptors at multiple spatial resolution levels generally increases the discriminative power of the descriptors. In this paper we focus on combination methods that combine multiple descriptors at multiple spatial resolution levels. A possible problem of the naive solution to create one large input vector for a machine learning classifier such as a support vector machine, is that the input vector becomes of very large dimensionality, which can increase problems of overfitting and hinder generalization performance. Therefore we propose the use of stacking support vector machines where at the first layer each support vector machine receives the input constructed by each single descriptor and is trained to compute the right output class. A second layer support vector machine is then used to combine the class probabilities of all trained first layer support vector models to learn the right output class given these reduced input vectors. We have performed experiments on 20 classes from the Caltech object database with 10 different single descriptors at 3 different resolutions. The results show that our 2-layer stacking approach outperforms the naive approach that combines all descriptors directly in a very large single input vector.",
author = "Azizi Abdullah and Veltkamp, {Remco C.} and Wiering, {Marco A.}",
year = "2009",
doi = "10.1109/IJCNN.2009.5178743",
language = "English",
isbn = "9781424435531",
pages = "5--12",
booktitle = "Proceedings of the International Joint Conference on Neural Networks",

}

TY - GEN

T1 - Spatial pyramids and two-layer stacking SVM classifiers for image categorization

T2 - A comparative study

AU - Abdullah, Azizi

AU - Veltkamp, Remco C.

AU - Wiering, Marco A.

PY - 2009

Y1 - 2009

N2 - Recent research in image recognition has shown that combining multiple descriptors is a very useful way to improve classification performance. Furthermore, the use of spatial pyramids that compute descriptors at multiple spatial resolution levels generally increases the discriminative power of the descriptors. In this paper we focus on combination methods that combine multiple descriptors at multiple spatial resolution levels. A possible problem of the naive solution to create one large input vector for a machine learning classifier such as a support vector machine, is that the input vector becomes of very large dimensionality, which can increase problems of overfitting and hinder generalization performance. Therefore we propose the use of stacking support vector machines where at the first layer each support vector machine receives the input constructed by each single descriptor and is trained to compute the right output class. A second layer support vector machine is then used to combine the class probabilities of all trained first layer support vector models to learn the right output class given these reduced input vectors. We have performed experiments on 20 classes from the Caltech object database with 10 different single descriptors at 3 different resolutions. The results show that our 2-layer stacking approach outperforms the naive approach that combines all descriptors directly in a very large single input vector.

AB - Recent research in image recognition has shown that combining multiple descriptors is a very useful way to improve classification performance. Furthermore, the use of spatial pyramids that compute descriptors at multiple spatial resolution levels generally increases the discriminative power of the descriptors. In this paper we focus on combination methods that combine multiple descriptors at multiple spatial resolution levels. A possible problem of the naive solution to create one large input vector for a machine learning classifier such as a support vector machine, is that the input vector becomes of very large dimensionality, which can increase problems of overfitting and hinder generalization performance. Therefore we propose the use of stacking support vector machines where at the first layer each support vector machine receives the input constructed by each single descriptor and is trained to compute the right output class. A second layer support vector machine is then used to combine the class probabilities of all trained first layer support vector models to learn the right output class given these reduced input vectors. We have performed experiments on 20 classes from the Caltech object database with 10 different single descriptors at 3 different resolutions. The results show that our 2-layer stacking approach outperforms the naive approach that combines all descriptors directly in a very large single input vector.

UR - http://www.scopus.com/inward/record.url?scp=70449399643&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=70449399643&partnerID=8YFLogxK

U2 - 10.1109/IJCNN.2009.5178743

DO - 10.1109/IJCNN.2009.5178743

M3 - Conference contribution

AN - SCOPUS:70449399643

SN - 9781424435531

SP - 5

EP - 12

BT - Proceedings of the International Joint Conference on Neural Networks

ER -