Large multi-class image categorization with ensembles of label trees

Yang Wang, David Forsyth

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

We consider sublinear test-time algorithms for image categorization when the number of classes is very large. Our method builds upon the label tree approach proposed in [1], which decomposes the label set into a tree structure and classify a test example by traversing the tree. Even though this method achieves logarithmic run-time, its performance is limited by the fact that any errors made in an internal node of the tree cannot be recovered. In this paper, we propose label forests - ensembles of label trees. Each tree in a label forest will decompose the label set in a slightly different way. The final classification decision is made by aggregating information across all trees in the label forest. The test running time of label forest is still logarithmic in the number of categories. But using an ensemble of label trees achieves much better performance in terms of accuracies. We demonstrate our approach on an image classification task that involves 1000 categories.

Original languageEnglish (US)
Title of host publication2013 IEEE International Conference on Multimedia and Expo, ICME 2013
DOIs
StatePublished - Oct 21 2013
Event2013 IEEE International Conference on Multimedia and Expo, ICME 2013 - San Jose, CA, United States
Duration: Jul 15 2013Jul 19 2013

Publication series

NameProceedings - IEEE International Conference on Multimedia and Expo
ISSN (Print)1945-7871
ISSN (Electronic)1945-788X

Other

Other2013 IEEE International Conference on Multimedia and Expo, ICME 2013
CountryUnited States
CitySan Jose, CA
Period7/15/137/19/13

Fingerprint

Labels
Image classification

Keywords

  • Image classification
  • large-scale learning

ASJC Scopus subject areas

  • Computer Networks and Communications
  • Computer Science Applications

Cite this

Wang, Y., & Forsyth, D. (2013). Large multi-class image categorization with ensembles of label trees. In 2013 IEEE International Conference on Multimedia and Expo, ICME 2013 [6607437] (Proceedings - IEEE International Conference on Multimedia and Expo). https://doi.org/10.1109/ICME.2013.6607437

Large multi-class image categorization with ensembles of label trees. / Wang, Yang; Forsyth, David.

2013 IEEE International Conference on Multimedia and Expo, ICME 2013. 2013. 6607437 (Proceedings - IEEE International Conference on Multimedia and Expo).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Wang, Y & Forsyth, D 2013, Large multi-class image categorization with ensembles of label trees. in 2013 IEEE International Conference on Multimedia and Expo, ICME 2013., 6607437, Proceedings - IEEE International Conference on Multimedia and Expo, 2013 IEEE International Conference on Multimedia and Expo, ICME 2013, San Jose, CA, United States, 7/15/13. https://doi.org/10.1109/ICME.2013.6607437
Wang Y, Forsyth D. Large multi-class image categorization with ensembles of label trees. In 2013 IEEE International Conference on Multimedia and Expo, ICME 2013. 2013. 6607437. (Proceedings - IEEE International Conference on Multimedia and Expo). https://doi.org/10.1109/ICME.2013.6607437
Wang, Yang ; Forsyth, David. / Large multi-class image categorization with ensembles of label trees. 2013 IEEE International Conference on Multimedia and Expo, ICME 2013. 2013. (Proceedings - IEEE International Conference on Multimedia and Expo).
@inproceedings{1bf639c8c47b4922a5cd206909225dc2,
title = "Large multi-class image categorization with ensembles of label trees",
abstract = "We consider sublinear test-time algorithms for image categorization when the number of classes is very large. Our method builds upon the label tree approach proposed in [1], which decomposes the label set into a tree structure and classify a test example by traversing the tree. Even though this method achieves logarithmic run-time, its performance is limited by the fact that any errors made in an internal node of the tree cannot be recovered. In this paper, we propose label forests - ensembles of label trees. Each tree in a label forest will decompose the label set in a slightly different way. The final classification decision is made by aggregating information across all trees in the label forest. The test running time of label forest is still logarithmic in the number of categories. But using an ensemble of label trees achieves much better performance in terms of accuracies. We demonstrate our approach on an image classification task that involves 1000 categories.",
keywords = "Image classification, large-scale learning",
author = "Yang Wang and David Forsyth",
year = "2013",
month = "10",
day = "21",
doi = "10.1109/ICME.2013.6607437",
language = "English (US)",
isbn = "9781479900152",
series = "Proceedings - IEEE International Conference on Multimedia and Expo",
booktitle = "2013 IEEE International Conference on Multimedia and Expo, ICME 2013",

}

TY - GEN

T1 - Large multi-class image categorization with ensembles of label trees

AU - Wang, Yang

AU - Forsyth, David

PY - 2013/10/21

Y1 - 2013/10/21

N2 - We consider sublinear test-time algorithms for image categorization when the number of classes is very large. Our method builds upon the label tree approach proposed in [1], which decomposes the label set into a tree structure and classify a test example by traversing the tree. Even though this method achieves logarithmic run-time, its performance is limited by the fact that any errors made in an internal node of the tree cannot be recovered. In this paper, we propose label forests - ensembles of label trees. Each tree in a label forest will decompose the label set in a slightly different way. The final classification decision is made by aggregating information across all trees in the label forest. The test running time of label forest is still logarithmic in the number of categories. But using an ensemble of label trees achieves much better performance in terms of accuracies. We demonstrate our approach on an image classification task that involves 1000 categories.

AB - We consider sublinear test-time algorithms for image categorization when the number of classes is very large. Our method builds upon the label tree approach proposed in [1], which decomposes the label set into a tree structure and classify a test example by traversing the tree. Even though this method achieves logarithmic run-time, its performance is limited by the fact that any errors made in an internal node of the tree cannot be recovered. In this paper, we propose label forests - ensembles of label trees. Each tree in a label forest will decompose the label set in a slightly different way. The final classification decision is made by aggregating information across all trees in the label forest. The test running time of label forest is still logarithmic in the number of categories. But using an ensemble of label trees achieves much better performance in terms of accuracies. We demonstrate our approach on an image classification task that involves 1000 categories.

KW - Image classification

KW - large-scale learning

UR - http://www.scopus.com/inward/record.url?scp=84885670722&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84885670722&partnerID=8YFLogxK

U2 - 10.1109/ICME.2013.6607437

DO - 10.1109/ICME.2013.6607437

M3 - Conference contribution

AN - SCOPUS:84885670722

SN - 9781479900152

T3 - Proceedings - IEEE International Conference on Multimedia and Expo

BT - 2013 IEEE International Conference on Multimedia and Expo, ICME 2013

ER -