Detail publikace

Handwritten Digits Recognition Improved by Multiresolution Classifier Fusion

ŠTRBA, M. HEROUT, A. HAVEL, J.

Originální název

Handwritten Digits Recognition Improved by Multiresolution Classifier Fusion

Anglický název

Handwritten Digits Recognition Improved by Multiresolution Classifier Fusion

Jazyk

en

Originální abstrakt

One common approach to construction of highly accurate classifiers for hadwritten digit recognition is fusion of several weaker classifiers into a compound one, which (when meeting some constraints) outperforms all the individual fused classifiers.  This paper studies the possibility of fusing classifiers of different kinds (Self-Organizing Maps, Randomized Trees, and AdaBoost with MB-LBP weak hypotheses) constructed on training sets resampled to different resolutions.  While it is common to select one resolution of the input samples as the ``ideal one'' and fuse classifiers constructed for it, this paper shows that the accuracy of classification can be improved by fusing information from several scales.

Anglický abstrakt

One common approach to construction of highly accurate classifiers for hadwritten digit recognition is fusion of several weaker classifiers into a compound one, which (when meeting some constraints) outperforms all the individual fused classifiers.  This paper studies the possibility of fusing classifiers of different kinds (Self-Organizing Maps, Randomized Trees, and AdaBoost with MB-LBP weak hypotheses) constructed on training sets resampled to different resolutions.  While it is common to select one resolution of the input samples as the ``ideal one'' and fuse classifiers constructed for it, this paper shows that the accuracy of classification can be improved by fusing information from several scales.

Dokumenty

BibTex


@inproceedings{BUT76284,
  author="Miroslav {Štrba} and Adam {Herout} and Jiří {Havel}",
  title="Handwritten Digits Recognition Improved by Multiresolution Classifier Fusion",
  annote="One common approach to construction of highly accurate classifiers for hadwritten
digit recognition is fusion of several weaker classifiers into a compound one,
which (when meeting some constraints) outperforms all the individual fused
classifiers.  This paper studies the possibility of fusing classifiers of
different kinds (Self-Organizing Maps, Randomized Trees, and AdaBoost with MB-LBP
weak hypotheses) constructed on training sets resampled to different
resolutions.  While it is common to select one resolution of the input samples as
the ``ideal one'' and fuse classifiers constructed for it, this paper shows that
the accuracy of classification can be improved by fusing information from several
scales.",
  address="Springer Verlag",
  booktitle="Proceedings of IbPRIA 2011, LNCS",
  chapter="76284",
  edition="NEUVEDEN",
  howpublished="print",
  institution="Springer Verlag",
  year="2011",
  month="june",
  pages="726--733",
  publisher="Springer Verlag",
  type="conference paper"
}