Semantic label sharing for learning with many categories

Rob Fergus; Hector Bernal; Yair Weiss; Antonio Torralba

doi:10.1007/978-3-642-15549-9_55

Semantic label sharing for learning with many categories

Rob Fergus^*, Hector Bernal, Yair Weiss, Antonio Torralba

^*Corresponding author for this work

The Rachel and Selim Benin School of Engineering and Computer Science

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

60 Scopus citations

Abstract

In an object recognition scenario with tens of thousands of categories, even a small number of labels per category leads to a very large number of total labels required. We propose a simple method of label sharing between semantically similar categories. We leverage the WordNet hierarchy to define semantic distance between any two categories and use this semantic distance to share labels. Our approach can be used with any classifier. Experimental results on a range of datasets, upto 80 million images and 75,000 categories in size, show that despite the simplicity of the approach, it leads to significant improvements in performance.

Original language	English
Title of host publication	Computer Vision, ECCV 2010 - 11th European Conference on Computer Vision, Proceedings
Publisher	Springer Verlag
Pages	762-775
Number of pages	14
Edition	PART 1
ISBN (Print)	3642155480, 9783642155482
DOIs	https://doi.org/10.1007/978-3-642-15549-9_55
State	Published - 2010
Event	11th European Conference on Computer Vision, ECCV 2010 - Heraklion, Crete, Greece Duration: 10 Sep 2010 → 11 Sep 2010

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Number	PART 1
Volume	6311 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	11th European Conference on Computer Vision, ECCV 2010
Country/Territory	Greece
City	Heraklion, Crete
Period	10/09/10 → 11/09/10

Access to Document

10.1007/978-3-642-15549-9_55

Cite this

Fergus, R., Bernal, H., Weiss, Y., & Torralba, A. (2010). Semantic label sharing for learning with many categories. In Computer Vision, ECCV 2010 - 11th European Conference on Computer Vision, Proceedings (PART 1 ed., pp. 762-775). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 6311 LNCS, No. PART 1). Springer Verlag. https://doi.org/10.1007/978-3-642-15549-9_55

Fergus, Rob ; Bernal, Hector ; Weiss, Yair et al. / Semantic label sharing for learning with many categories. Computer Vision, ECCV 2010 - 11th European Conference on Computer Vision, Proceedings. PART 1. ed. Springer Verlag, 2010. pp. 762-775 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); PART 1).

@inproceedings{3bae44b193ca4788a5564dfdc283e41a,

title = "Semantic label sharing for learning with many categories",

abstract = "In an object recognition scenario with tens of thousands of categories, even a small number of labels per category leads to a very large number of total labels required. We propose a simple method of label sharing between semantically similar categories. We leverage the WordNet hierarchy to define semantic distance between any two categories and use this semantic distance to share labels. Our approach can be used with any classifier. Experimental results on a range of datasets, upto 80 million images and 75,000 categories in size, show that despite the simplicity of the approach, it leads to significant improvements in performance.",

author = "Rob Fergus and Hector Bernal and Yair Weiss and Antonio Torralba",

year = "2010",

doi = "10.1007/978-3-642-15549-9_55",

language = "אנגלית",

isbn = "3642155480",

series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",

publisher = "Springer Verlag",

number = "PART 1",

pages = "762--775",

booktitle = "Computer Vision, ECCV 2010 - 11th European Conference on Computer Vision, Proceedings",

address = "גרמניה",

edition = "PART 1",

note = "11th European Conference on Computer Vision, ECCV 2010 ; Conference date: 10-09-2010 Through 11-09-2010",

}

Fergus, R, Bernal, H, Weiss, Y & Torralba, A 2010, Semantic label sharing for learning with many categories. in Computer Vision, ECCV 2010 - 11th European Conference on Computer Vision, Proceedings. PART 1 edn, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), no. PART 1, vol. 6311 LNCS, Springer Verlag, pp. 762-775, 11th European Conference on Computer Vision, ECCV 2010, Heraklion, Crete, Greece, 10/09/10. https://doi.org/10.1007/978-3-642-15549-9_55

Semantic label sharing for learning with many categories. / Fergus, Rob; Bernal, Hector; Weiss, Yair et al.
Computer Vision, ECCV 2010 - 11th European Conference on Computer Vision, Proceedings. PART 1. ed. Springer Verlag, 2010. p. 762-775 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 6311 LNCS, No. PART 1).

Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › peer-review

TY - GEN

T1 - Semantic label sharing for learning with many categories

AU - Fergus, Rob

AU - Bernal, Hector

AU - Weiss, Yair

AU - Torralba, Antonio

PY - 2010

Y1 - 2010

N2 - In an object recognition scenario with tens of thousands of categories, even a small number of labels per category leads to a very large number of total labels required. We propose a simple method of label sharing between semantically similar categories. We leverage the WordNet hierarchy to define semantic distance between any two categories and use this semantic distance to share labels. Our approach can be used with any classifier. Experimental results on a range of datasets, upto 80 million images and 75,000 categories in size, show that despite the simplicity of the approach, it leads to significant improvements in performance.

AB - In an object recognition scenario with tens of thousands of categories, even a small number of labels per category leads to a very large number of total labels required. We propose a simple method of label sharing between semantically similar categories. We leverage the WordNet hierarchy to define semantic distance between any two categories and use this semantic distance to share labels. Our approach can be used with any classifier. Experimental results on a range of datasets, upto 80 million images and 75,000 categories in size, show that despite the simplicity of the approach, it leads to significant improvements in performance.

UR - http://www.scopus.com/inward/record.url?scp=78149329142&partnerID=8YFLogxK

U2 - 10.1007/978-3-642-15549-9_55

DO - 10.1007/978-3-642-15549-9_55

M3 - ???researchoutput.researchoutputtypes.contributiontobookanthology.conference???

AN - SCOPUS:78149329142

SN - 3642155480

SN - 9783642155482

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

SP - 762

EP - 775

BT - Computer Vision, ECCV 2010 - 11th European Conference on Computer Vision, Proceedings

PB - Springer Verlag

T2 - 11th European Conference on Computer Vision, ECCV 2010

Y2 - 10 September 2010 through 11 September 2010

ER -

Fergus R, Bernal H, Weiss Y, Torralba A. Semantic label sharing for learning with many categories. In Computer Vision, ECCV 2010 - 11th European Conference on Computer Vision, Proceedings. PART 1 ed. Springer Verlag. 2010. p. 762-775. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); PART 1). doi: 10.1007/978-3-642-15549-9_55

Semantic label sharing for learning with many categories

Abstract

Publication series

Conference

Access to Document

Other files and links

Fingerprint

Cite this