Attribute discovery via predictable discriminative binary codes

Mohammad Rastegari, Ali Farhadi, David Forsyth

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

Abstract

We present images with binary codes in a way that balances discrimination and learnability of the codes. In our method, each image claims its own code in a way that maintains discrimination while being predictable from visual data. Category memberships are usually good proxies for visual similarity but should not be enforced as a hard constraint. Our method learns codes that maximize separability of categories unless there is strong visual evidence against it. Simple linear SVMs can achieve state-of-the-art results with our short codes. In fact, our method produces state-of-the-art results on Caltech256 with only 128-dimensional bit vectors and outperforms state of the art by using longer codes. We also evaluate our method on ImageNet and show that our method outperforms state-of-the-art binary code methods on this large scale dataset. Lastly, our codes can discover a discriminative set of attributes.

Original languageEnglish (US)
Title of host publicationComputer Vision, ECCV 2012 - 12th European Conference on Computer Vision, Proceedings
Pages876-889
Number of pages14
EditionPART 6
DOIs
StatePublished - Oct 30 2012
Event12th European Conference on Computer Vision, ECCV 2012 - Florence, Italy
Duration: Oct 7 2012Oct 13 2012

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
NumberPART 6
Volume7577 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Other

Other12th European Conference on Computer Vision, ECCV 2012
CountryItaly
CityFlorence
Period10/7/1210/13/12

Fingerprint

Binary codes
Binary Code
Attribute
Discrimination
Learnability
Separability
Maximise
Evaluate
Vision

ASJC Scopus subject areas

  • Theoretical Computer Science
  • Computer Science(all)

Cite this

Rastegari, M., Farhadi, A., & Forsyth, D. (2012). Attribute discovery via predictable discriminative binary codes. In Computer Vision, ECCV 2012 - 12th European Conference on Computer Vision, Proceedings (PART 6 ed., pp. 876-889). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 7577 LNCS, No. PART 6). https://doi.org/10.1007/978-3-642-33783-3_63

Attribute discovery via predictable discriminative binary codes. / Rastegari, Mohammad; Farhadi, Ali; Forsyth, David.

Computer Vision, ECCV 2012 - 12th European Conference on Computer Vision, Proceedings. PART 6. ed. 2012. p. 876-889 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 7577 LNCS, No. PART 6).

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

Rastegari, M, Farhadi, A & Forsyth, D 2012, Attribute discovery via predictable discriminative binary codes. in Computer Vision, ECCV 2012 - 12th European Conference on Computer Vision, Proceedings. PART 6 edn, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), no. PART 6, vol. 7577 LNCS, pp. 876-889, 12th European Conference on Computer Vision, ECCV 2012, Florence, Italy, 10/7/12. https://doi.org/10.1007/978-3-642-33783-3_63
Rastegari M, Farhadi A, Forsyth D. Attribute discovery via predictable discriminative binary codes. In Computer Vision, ECCV 2012 - 12th European Conference on Computer Vision, Proceedings. PART 6 ed. 2012. p. 876-889. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); PART 6). https://doi.org/10.1007/978-3-642-33783-3_63
Rastegari, Mohammad ; Farhadi, Ali ; Forsyth, David. / Attribute discovery via predictable discriminative binary codes. Computer Vision, ECCV 2012 - 12th European Conference on Computer Vision, Proceedings. PART 6. ed. 2012. pp. 876-889 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); PART 6).
@inproceedings{0a65edbfbe0e4027ae32d58d6bab6108,
title = "Attribute discovery via predictable discriminative binary codes",
abstract = "We present images with binary codes in a way that balances discrimination and learnability of the codes. In our method, each image claims its own code in a way that maintains discrimination while being predictable from visual data. Category memberships are usually good proxies for visual similarity but should not be enforced as a hard constraint. Our method learns codes that maximize separability of categories unless there is strong visual evidence against it. Simple linear SVMs can achieve state-of-the-art results with our short codes. In fact, our method produces state-of-the-art results on Caltech256 with only 128-dimensional bit vectors and outperforms state of the art by using longer codes. We also evaluate our method on ImageNet and show that our method outperforms state-of-the-art binary code methods on this large scale dataset. Lastly, our codes can discover a discriminative set of attributes.",
author = "Mohammad Rastegari and Ali Farhadi and David Forsyth",
year = "2012",
month = "10",
day = "30",
doi = "10.1007/978-3-642-33783-3_63",
language = "English (US)",
isbn = "9783642337826",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
number = "PART 6",
pages = "876--889",
booktitle = "Computer Vision, ECCV 2012 - 12th European Conference on Computer Vision, Proceedings",
edition = "PART 6",

}

TY - GEN

T1 - Attribute discovery via predictable discriminative binary codes

AU - Rastegari, Mohammad

AU - Farhadi, Ali

AU - Forsyth, David

PY - 2012/10/30

Y1 - 2012/10/30

N2 - We present images with binary codes in a way that balances discrimination and learnability of the codes. In our method, each image claims its own code in a way that maintains discrimination while being predictable from visual data. Category memberships are usually good proxies for visual similarity but should not be enforced as a hard constraint. Our method learns codes that maximize separability of categories unless there is strong visual evidence against it. Simple linear SVMs can achieve state-of-the-art results with our short codes. In fact, our method produces state-of-the-art results on Caltech256 with only 128-dimensional bit vectors and outperforms state of the art by using longer codes. We also evaluate our method on ImageNet and show that our method outperforms state-of-the-art binary code methods on this large scale dataset. Lastly, our codes can discover a discriminative set of attributes.

AB - We present images with binary codes in a way that balances discrimination and learnability of the codes. In our method, each image claims its own code in a way that maintains discrimination while being predictable from visual data. Category memberships are usually good proxies for visual similarity but should not be enforced as a hard constraint. Our method learns codes that maximize separability of categories unless there is strong visual evidence against it. Simple linear SVMs can achieve state-of-the-art results with our short codes. In fact, our method produces state-of-the-art results on Caltech256 with only 128-dimensional bit vectors and outperforms state of the art by using longer codes. We also evaluate our method on ImageNet and show that our method outperforms state-of-the-art binary code methods on this large scale dataset. Lastly, our codes can discover a discriminative set of attributes.

UR - http://www.scopus.com/inward/record.url?scp=84867889923&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=84867889923&partnerID=8YFLogxK

U2 - 10.1007/978-3-642-33783-3_63

DO - 10.1007/978-3-642-33783-3_63

M3 - Conference contribution

AN - SCOPUS:84867889923

SN - 9783642337826

T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)

SP - 876

EP - 889

BT - Computer Vision, ECCV 2012 - 12th European Conference on Computer Vision, Proceedings

ER -