Object Discovery via Contrastive Learning for Weakly Supervised Object Detection

Jinhwan Seo; Wonho Bae; Danica J. Sutherland; Junhyug Noh; Daijin Kim

doi:10.1007/978-3-031-19821-2_18

Object Discovery via Contrastive Learning for Weakly Supervised Object Detection

Jinhwan Seo, Wonho Bae, Danica J. Sutherland, Junhyug Noh, Daijin Kim

Department of Artificial Intelligence

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

17 Scopus citations

Abstract

Weakly Supervised Object Detection (WSOD) is a task that detects objects in an image using a model trained only on image-level annotations. Current state-of-the-art models benefit from self-supervised instance-level supervision, but since weak supervision does not include count or location information, the most common “argmax” labeling method often ignores many instances of objects. To alleviate this issue, we propose a novel multiple instance labeling method called object discovery. We further introduce a new contrastive loss under weak supervision where no instance-level information is available for sampling, called weakly supervised contrastive loss (WSCL). WSCL aims to construct a credible similarity threshold for object discovery by leveraging consistent features for embedding vectors in the same class. As a result, we achieve new state-of-the-art results on MS-COCO 2014 and 2017 as well as PASCAL VOC 2012, and competitive results on PASCAL VOC 2007. The code is available at https://github.com/jinhseo/OD-WSCL.

Original language	English
Title of host publication	Computer Vision – ECCV 2022 - 17th European Conference, Proceedings
Editors	Shai Avidan, Gabriel Brostow, Moustapha Cissé, Giovanni Maria Farinella, Tal Hassner
Publisher	Springer Science and Business Media Deutschland GmbH
Pages	312-329
Number of pages	18
ISBN (Print)	9783031198205
DOIs	https://doi.org/10.1007/978-3-031-19821-2_18
State	Published - 2022
Event	17th European Conference on Computer Vision, ECCV 2022 - Tel Aviv, Israel Duration: 23 Oct 2022 → 27 Oct 2022

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	13691 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	17th European Conference on Computer Vision, ECCV 2022
Country/Territory	Israel
City	Tel Aviv
Period	23/10/22 → 27/10/22

Bibliographical note

Publisher Copyright:
© 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.

Keywords

Weakly Supervised Object Detection (WSOD)

Access to Document

10.1007/978-3-031-19821-2_18

Cite this

Seo, J., Bae, W., Sutherland, D. J., Noh, J., & Kim, D. (2022). Object Discovery via Contrastive Learning for Weakly Supervised Object Detection. In S. Avidan, G. Brostow, M. Cissé, G. M. Farinella, & T. Hassner (Eds.), Computer Vision – ECCV 2022 - 17th European Conference, Proceedings (pp. 312-329). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 13691 LNCS). Springer Science and Business Media Deutschland GmbH. https://doi.org/10.1007/978-3-031-19821-2_18

Seo, Jinhwan ; Bae, Wonho ; Sutherland, Danica J. et al. / Object Discovery via Contrastive Learning for Weakly Supervised Object Detection. Computer Vision – ECCV 2022 - 17th European Conference, Proceedings. editor / Shai Avidan ; Gabriel Brostow ; Moustapha Cissé ; Giovanni Maria Farinella ; Tal Hassner. Springer Science and Business Media Deutschland GmbH, 2022. pp. 312-329 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{c66412194b3d4ff5bcce49c88c9e23a7,

title = "Object Discovery via Contrastive Learning for Weakly Supervised Object Detection",

abstract = "Weakly Supervised Object Detection (WSOD) is a task that detects objects in an image using a model trained only on image-level annotations. Current state-of-the-art models benefit from self-supervised instance-level supervision, but since weak supervision does not include count or location information, the most common “argmax” labeling method often ignores many instances of objects. To alleviate this issue, we propose a novel multiple instance labeling method called object discovery. We further introduce a new contrastive loss under weak supervision where no instance-level information is available for sampling, called weakly supervised contrastive loss (WSCL). WSCL aims to construct a credible similarity threshold for object discovery by leveraging consistent features for embedding vectors in the same class. As a result, we achieve new state-of-the-art results on MS-COCO 2014 and 2017 as well as PASCAL VOC 2012, and competitive results on PASCAL VOC 2007. The code is available at https://github.com/jinhseo/OD-WSCL.",

keywords = "Weakly Supervised Object Detection (WSOD)",

author = "Jinhwan Seo and Wonho Bae and Sutherland, {Danica J.} and Junhyug Noh and Daijin Kim",

note = "Publisher Copyright: {\textcopyright} 2022, The Author(s), under exclusive license to Springer Nature Switzerland AG.; 17th European Conference on Computer Vision, ECCV 2022 ; Conference date: 23-10-2022 Through 27-10-2022",

year = "2022",

doi = "10.1007/978-3-031-19821-2_18",

language = "English",

isbn = "9783031198205",

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

publisher = "Springer Science and Business Media Deutschland GmbH",

pages = "312--329",

editor = "Shai Avidan and Gabriel Brostow and Moustapha Ciss{\'e} and Farinella, {Giovanni Maria} and Tal Hassner",

booktitle = "Computer Vision – ECCV 2022 - 17th European Conference, Proceedings",

}

Seo, J, Bae, W, Sutherland, DJ, Noh, J & Kim, D 2022, Object Discovery via Contrastive Learning for Weakly Supervised Object Detection. in S Avidan, G Brostow, M Cissé, GM Farinella & T Hassner (eds), Computer Vision – ECCV 2022 - 17th European Conference, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 13691 LNCS, Springer Science and Business Media Deutschland GmbH, pp. 312-329, 17th European Conference on Computer Vision, ECCV 2022, Tel Aviv, Israel, 23/10/22. https://doi.org/10.1007/978-3-031-19821-2_18

Object Discovery via Contrastive Learning for Weakly Supervised Object Detection. / Seo, Jinhwan; Bae, Wonho; Sutherland, Danica J. et al.
Computer Vision – ECCV 2022 - 17th European Conference, Proceedings. ed. / Shai Avidan; Gabriel Brostow; Moustapha Cissé; Giovanni Maria Farinella; Tal Hassner. Springer Science and Business Media Deutschland GmbH, 2022. p. 312-329 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 13691 LNCS).

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

TY - GEN

T1 - Object Discovery via Contrastive Learning for Weakly Supervised Object Detection

AU - Seo, Jinhwan

AU - Bae, Wonho

AU - Sutherland, Danica J.

AU - Noh, Junhyug

AU - Kim, Daijin

PY - 2022

Y1 - 2022

N2 - Weakly Supervised Object Detection (WSOD) is a task that detects objects in an image using a model trained only on image-level annotations. Current state-of-the-art models benefit from self-supervised instance-level supervision, but since weak supervision does not include count or location information, the most common “argmax” labeling method often ignores many instances of objects. To alleviate this issue, we propose a novel multiple instance labeling method called object discovery. We further introduce a new contrastive loss under weak supervision where no instance-level information is available for sampling, called weakly supervised contrastive loss (WSCL). WSCL aims to construct a credible similarity threshold for object discovery by leveraging consistent features for embedding vectors in the same class. As a result, we achieve new state-of-the-art results on MS-COCO 2014 and 2017 as well as PASCAL VOC 2012, and competitive results on PASCAL VOC 2007. The code is available at https://github.com/jinhseo/OD-WSCL.

AB - Weakly Supervised Object Detection (WSOD) is a task that detects objects in an image using a model trained only on image-level annotations. Current state-of-the-art models benefit from self-supervised instance-level supervision, but since weak supervision does not include count or location information, the most common “argmax” labeling method often ignores many instances of objects. To alleviate this issue, we propose a novel multiple instance labeling method called object discovery. We further introduce a new contrastive loss under weak supervision where no instance-level information is available for sampling, called weakly supervised contrastive loss (WSCL). WSCL aims to construct a credible similarity threshold for object discovery by leveraging consistent features for embedding vectors in the same class. As a result, we achieve new state-of-the-art results on MS-COCO 2014 and 2017 as well as PASCAL VOC 2012, and competitive results on PASCAL VOC 2007. The code is available at https://github.com/jinhseo/OD-WSCL.

KW - Weakly Supervised Object Detection (WSOD)

UR - http://www.scopus.com/inward/record.url?scp=85142767029&partnerID=8YFLogxK

U2 - 10.1007/978-3-031-19821-2_18

DO - 10.1007/978-3-031-19821-2_18

M3 - Conference contribution

AN - SCOPUS:85142767029

SN - 9783031198205

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

SP - 312

EP - 329

BT - Computer Vision – ECCV 2022 - 17th European Conference, Proceedings

A2 - Avidan, Shai

A2 - Brostow, Gabriel

A2 - Cissé, Moustapha

A2 - Farinella, Giovanni Maria

A2 - Hassner, Tal

PB - Springer Science and Business Media Deutschland GmbH

T2 - 17th European Conference on Computer Vision, ECCV 2022

Y2 - 23 October 2022 through 27 October 2022

ER -

Seo J, Bae W, Sutherland DJ, Noh J, Kim D. Object Discovery via Contrastive Learning for Weakly Supervised Object Detection. In Avidan S, Brostow G, Cissé M, Farinella GM, Hassner T, editors, Computer Vision – ECCV 2022 - 17th European Conference, Proceedings. Springer Science and Business Media Deutschland GmbH. 2022. p. 312-329. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-031-19821-2_18