Greedy-Based Design of Sparse Two-Stage SVMs for Fast Classification

  • Rezaul Karim
  • Martin Bergtholdt
  • Jörg Kappes
  • Christoph Schnörr
Part of the Lecture Notes in Computer Science book series (LNCS, volume 4713)


Cascades of classifiers constitute an important architecture for fast object detection. While boosting of simple (weak) classifiers provides an established framework, the design of similar architectures with more powerful (strong) classifiers has become the subject of current research. In this paper, we focus on greedy strategies recently proposed in the literature that allow to learn sparse Support Vector Machines (SVMs) without the need to train full SVMs beforehand. We show (i) that asymmetric data sets that are typical for object detection scenarios can be successfully handled, and (ii) that the complementary training of two sparse SVMs leads to sequential two-stage classifiers that slightly outperform a full SVM, but only need about 10% kernel evaluations for classifying a pattern.


Support Vector Machine Support Vector Face Detection Greedy Search Greedy Strategy 
These keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.


Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.


  1. 1.
    Keerthi, S.S., Chapelle, O., DeCoste, D.: Building support vector machines with reduced classifier complexity. J. Mach. Learning Res. 7, 1493–1515 (2006)MathSciNetGoogle Scholar
  2. 2.
  3. 3.
    Bach, F.R., Heckerman, D., Horvitz, E.: Considering cost asymmetry in learning classifiers. J. Mach. Learning Res. 7, 1713–1741 (2006)MathSciNetGoogle Scholar
  4. 4.
    Viola, P., Jones, M.J.: Robust real-time face detection. Int. J. Comp. Vision 57(2), 137–154 (2004)Google Scholar
  5. 5.
    Romdhani, S., Torr, P., Schölkopf, B., Blake, A.: Efficient face detection by a cascaded support-vector machine expansion. Proc. Royal Soc. A 460, 3283–3297 (2004)zbMATHGoogle Scholar
  6. 6.
    Franc, V., Hlavać, V.: Greedy algorithm for a training set reduction in the kernel methods. In: Petkov, N., Westenberg, M.A. (eds.) CAIP 2003. LNCS, vol. 2756, pp. 426–433. Springer, Heidelberg (2003)Google Scholar
  7. 7.
    Sahbi, H., Geman, D.: A hierarchy of support vector machines for pattern detection. J. Mach. Learning Res. 7, 2087–2123 (2006)MathSciNetGoogle Scholar
  8. 8.
    Rätsch, M., Romdhani, S., Teschke, G., Vetter, T.: Over-complete wavelet approximation of a support vector machine for efficient classification. In: Kropatsch, W.G., Sablatnig, R., Hanbury, A. (eds.) Pattern Recognition. LNCS, vol. 3663, pp. 351–360. Springer, Heidelberg (2005)Google Scholar
  9. 9.
    Zapién, K., Fehr, J., Burkhardt, H.: Fast support vector machine classification using linear SVMs. In: 18th International Conference on Pattern Recognition (ICPR 2006), vol. 3, pp. 366–369. IEEE, Los Alamitos (2006)Google Scholar
  10. 10.
    Heisele, B., Serre, T., Prentice, S., Poggio, T.: Hierarchical classification and feature reduction for fast face detection with support vector machines. Pattern Recognition 36(9), 2007–2017 (2003)zbMATHCrossRefGoogle Scholar
  11. 11.
    Wu, M., Schölkopf, B., Bakir, G.: A direct method for building sparse kernel learning algorithms. J. Mach. Learning Res. 7, 603–624 (2006)Google Scholar
  12. 12.
    Lowe, D.G.: Distinctive image features from scale-invariant keypoints. Int. J. Comp. Vision 60(2), 91–110 (2004)Google Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 2007

Authors and Affiliations

  • Rezaul Karim
    • 1
  • Martin Bergtholdt
    • 1
  • Jörg Kappes
    • 1
  • Christoph Schnörr
    • 1
  1. 1.University of Mannheim, Dept. Math & CS – CVGPR Group 

Personalised recommendations