Geo-temporal Structuring of a Personal Image Database with Two-Level Variational-Bayes Mixture Estimation

  • Pierrick Bruneau
  • Antoine Pigeau
  • Marc Gelgon
  • Fabien Picarougne
Conference paper

DOI: 10.1007/978-3-642-14758-6_11

Part of the Lecture Notes in Computer Science book series (LNCS, volume 5811)
Cite this paper as:
Bruneau P., Pigeau A., Gelgon M., Picarougne F. (2010) Geo-temporal Structuring of a Personal Image Database with Two-Level Variational-Bayes Mixture Estimation. In: Detyniecki M., Leiner U., Nürnberger A. (eds) Adaptive Multimedia Retrieval. Identifying, Summarizing, and Recommending Image and Music. AMR 2008. Lecture Notes in Computer Science, vol 5811. Springer, Berlin, Heidelberg

Abstract

This paper addresses unsupervised hierarchical classification of personal documents tagged with time and geolocation stamps. The target application is browsing among these documents. A first partition of the data is built, based on geo-temporal measurement. The events found are then grouped according to geolocation. This is carried out through fitting a two-level hierarchy of mixture models to the data. Both mixtures are estimated in a Bayesian setting, with a variational procedure: the classical VBEM algorithm is applied for the finer level, while a new variational-Bayes-EM algorithm is introduced to search for suitable groups of mixture components from the finer level. Experimental results are reported on artificial and real data.

Preview

Unable to display preview. Download preview PDF.

Unable to display preview. Download preview PDF.

Copyright information

© Springer-Verlag Berlin Heidelberg 2010

Authors and Affiliations

  • Pierrick Bruneau
    • 1
    • 2
  • Antoine Pigeau
    • 1
  • Marc Gelgon
    • 1
    • 2
  • Fabien Picarougne
    • 1
  1. 1.Nantes university, LINA (UMR CNRS 6241), Polytech’NantesNantes cedex 3France
  2. 2.INRIA/IRISA Atlas project-team 

Personalised recommendations