Building on the ideas of Viola-Jones [1] we present a framework for training cascades of boosted ensembles (CoBE) which introduces further modularity and tractability to the training process. It addresses the challenges faced by CoBE frameworks such as protracted runtimes, slow layer convergences and classifier optimization. The framework possesses the ability to bootstrap positive samples and may in turn be extended into the domain of incremental learning. This paper aims to address our framework’s susceptibility to overfitting with possible solutions. Experiments are conducted on face detectors using the bootstrapping of large positive datasets and their accuracy, with respect to overfitting, is examined.


cascades of boosted ensembles AdaBoost classification classifier training face detection 


  1. 1.
    Viola, P., Jones, M.: Rapid object detection using a boosted cascade of simple features. In: CVPR 2001, Kauai, HI, December 2001, vol. I, pp. 511–518. IEEE, Los Alamitos (2001)Google Scholar
  2. 2.
    Brubaker, S.C., Mullin, M.D., Rehg, J.M.: Towards optimal training of cascaded detectors. In: Leonardis, A., Bischof, H., Pinz, A. (eds.) ECCV 2006. LNCS, vol. 3951, pp. 325–337. Springer, Heidelberg (2006)CrossRefGoogle Scholar
  3. 3.
    Pham, M.T., Cham, T.J.: Fast training and selection of haar features using statistics in boosting-based face detection. In: IEEE 11th International Conference on Computer Vision, 2007, October 2007, pp. 1–7 (2007)Google Scholar
  4. 4.
    Brubaker, S.C., Wu, J., Sun, J., Mullin, M.D., Rehg, J.M.: On the design of cascades of boosted ensembles for face detection. Int. J. Comput. Vision 77(1-3), 65–86 (2008)CrossRefGoogle Scholar
  5. 5.
    Wu, J., Rehg, J.M., Mullin, M.D.: Learning a rare event detection cascade by direct feature selection. In: NIPS Advances in Neural Information Processing Systems 2003, Vancouver, Canada (2003)Google Scholar
  6. 6.
    Lienhart, R., Maydt, J.: An extended set of haar-like features for rapid object detection. In: ICIP 2002, Rochester, NY, September 2002, vol. I, pp. 900–903 (2002)Google Scholar
  7. 7.
    Viola, P.A., Jones, M.J.: Fast and robust classification using asymmetric adaboost and a detector cascade. In: Dietterich, T.G., Becker, S., Ghahramani, Z. (eds.) NIPS, pp. 1311–1318 (2001)Google Scholar
  8. 8.
    Xiao, R., Zhu, H., Sun, H., Tang, X.: Dynamic cascades for face detection. In: IEEE 11th International Conference on Computer Vision 2007, October 2007, pp. 1–8 (2007)Google Scholar
  9. 9.
    Yan, S., Shan, S., Chen, X., Gao, W., Chen, J.: Matrix-Structural Learning (MSL) of cascaded classifier from enormous training set (2007)Google Scholar
  10. 10.
    Huang, C., Ai, H., Yamashita, T., Lao, S., Kawade, M.: Incremental learning of boosted face detector. In: ICCV, pp. 1–8. IEEE, Los Alamitos (2007)Google Scholar
  11. 11.
    Barczak, A.L.C., Johnson, M.J., Messom, C.H.: Empirical evaluation of a new structure for adaboost. In: SAC 2008, pp. 1764–1765. ACM, New York (2008)CrossRefGoogle Scholar
  12. 12.
    Susnjak, T., Barczak, A.L.C., Hawick, K.A.: A novel bootstrapping method for positive datasets in cascades of boosted ensembles. Research Letters in the Information and Mathematical Sciences Vol. 14, pp.17-24, Institute of Information and Mathematical Sciences, Massey University Albany (2010) ISSN 1175-2777Google Scholar
  13. 13.
    Georghiades, A., Belhumeur, P., Kriegman, D.: From few to many: Illumination cone models for face recognition under variable lighting and pose. IEEE Trans. Pattern Anal. Mach. Intelligence 23(6), 643–660 (2001)CrossRefGoogle Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 2010

Authors and Affiliations

  • Teo Susnjak
    • 1
  • Andre L. Barczak
    • 1
  • Ken A. Hawick
    • 1
  1. 1.Institute of Information and Mathematical SciencesMassey UniversityAlbanyNew Zealand

Personalised recommendations