Towards Learning Hierarchical Compositional Models in the Presence of Clutter

  • Jan Mačák
  • Ondřej Drbohlav
Part of the Lecture Notes in Computer Science book series (LNCS, volume 8156)

Abstract

Our goal is to identify hierarchical compositional models from highly cluttered data. The data to learn from are assumed to be imperfect in two respects. Firstly, large portion of the data is coming from background clutter. Secondly, data generated by a recursive compositional model are subject to random replacements of correct descendants by randomly chosen ones at every level of the hierarchy. In this paper, we study the limits and capabilities of an approach which is based on likelihood maximization. The algorithm makes explicit probabilistic assignments of individual data to compositional model and background clutter. It uses these assignments to effectively focus on the data coming from the compositional model and iteratively estimate their compositional structure.

Keywords

Ground Truth Compositional Model Minimum Description Length Background Clutter Minimum Description Length Principle 
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.

References

  1. 1.
    Bienenstock, E., Geman, S., Potter, D.: Compositionality, MDL priors, and object recognition. In: Mozer, M., Jordan, M.I., Petsche, T. (eds.) NIPS, pp. 838–844. MIT Press (1996)Google Scholar
  2. 2.
    Fidler, S., Boben, M., Leonardis, A.: Optimization framework for learning a hierarchical shape vocabulary for object class. In: BMVC. British Machine Vision Association (2009)Google Scholar
  3. 3.
    Fidler, S., Leonardis, A.: Towards scalable representations of object categories: Learning a hierarchy of parts. In: Proc. CVPR (2007)Google Scholar
  4. 4.
    Frey, B., Dayan, P., Hinton, G.E., Jenkin, I.M.: A simple algorithm that discovers efficient perceptual codes. In: Mechanisms, L.R.H., Mechanisms, B. (ed.) Computational and Psychophysical Mechanisms of Visual Coding, pp. 296–315. Cambridge University Press (1997)Google Scholar
  5. 5.
    Fukushima, K.: Neocognitron: A hierarchical neural network capable of visual pattern recognition. Neural Networks 1(2), 119–130 (1988), http://www.sciencedirect.com/science/article/pii/0893608088900147 CrossRefGoogle Scholar
  6. 6.
    Grünwald, P.D.: The Minimum Description Length Principle. MIT Press (2007)Google Scholar
  7. 7.
    Hinton, G.E., Osindero, S.: A fast learning algorithm for deep belief nets. Neural Computation 18 (2006)CrossRefGoogle Scholar
  8. 8.
    Kokkinos, I., Yuille, A.: Inference and learning with hierarchical shape models. International Journal of Computer Vision 93, 201–225 (2011), http://dx.doi.org/10.1007/s11263-010-0398-7 MathSciNetCrossRefMATHGoogle Scholar
  9. 9.
    Ommer, B., Buhmann, J.: A compositionality architecture for perceptual feature grouping. In: Rangarajan, A., Figueiredo, M.A.T., Zerubia, J. (eds.) EMMCVPR 2003. LNCS, vol. 2683, pp. 275–290. Springer, Heidelberg (2003)CrossRefGoogle Scholar
  10. 10.
    Ommer, B., Buhmann, J.: Learning the compositional nature of visual object categories for recognition. IEEE Trans. Pattern Anal. Mach. Intell. 32(3), 501–516 (2010)CrossRefGoogle Scholar
  11. 11.
    Tsotsos, J.K.: Analyzing vision at the complexity level. Behavioral and Brain Sciences 13(03), 423–445 (1990), http://dx.doi.org/10.1017/S0140525X00079577 CrossRefGoogle Scholar
  12. 12.
    Yuille, A.L.: Towards a theory of compositional learning and encoding of objects. In: ICCV Workshops, pp. 1448–1455. IEEE (2011)Google Scholar
  13. 13.
    Zhu, L., Chen, Y., Yuille, A.L.: Learning a hierarchical deformable template for rapid deformable object parsing. IEEE Trans. Pattern Anal. Mach. Intell. 32(6), 1029–1043 (2010)CrossRefGoogle Scholar
  14. 14.
    Zhu, L., Chen, Y., Yuille, A.L.: Recursive compositional models for vision: Description and review of recent work. Journal of Mathematical Imaging and Vision 41(1-2), 122–146 (2011)MathSciNetCrossRefMATHGoogle Scholar

Copyright information

© Springer-Verlag Berlin Heidelberg 2013

Authors and Affiliations

  • Jan Mačák
    • 1
  • Ondřej Drbohlav
    • 1
  1. 1.Center for Machine Perception (CMP), Department of CyberneticsCzech Technical University in PragueCzech Republic

Personalised recommendations