Detecting people in artwork with CNNs

Nicholas Westlake; Hongping Cai; Peter Hall

doi:10.1007/978-3-319-46604-0_57

Detecting people in artwork with CNNs

Nicholas Westlake, Hongping Cai, Peter Hall

Research output: Chapter in Book/Report/Conference proceeding › Chapter in a published conference proceeding

20 Citations (SciVal)

Abstract

CNNs have massively improved performance in object detection in photographs. However research into object detection in artwork remains limited. We show state-of-the-art performance on a challenging dataset, People-Art, which contains people from photos, cartoons and 41 different artwork movements. We achieve this high performance by finetuning a CNN for this task, thus also demonstrating that training CNNs on photos results in overfitting for photos: only the first three or four layers transfer from photos to artwork. Although the CNN’s performance is the highest yet, it remains less than 60% AP, suggesting further work is needed for the cross-depiction problem.

Original language	English
Title of host publication	Computer Vision - ECCV 2016 Workshops. ECCV 2016.
Editors	Gang Hua, Hervé Jégou
Publisher	Springer Verlag
Pages	825-841
Number of pages	17
ISBN (Print)	9783319466033
DOIs	https://doi.org/10.1007/978-3-319-46604-0_57
Publication status	Published - 18 Sep 2016
Event	14th European Conference on Computer Vision, ECCV 2016 - Amsterdam, Netherlands Duration: 8 Oct 2016 → 16 Oct 2016

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	9913 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	14th European Conference on Computer Vision, ECCV 2016
Country/Territory	Netherlands
City	Amsterdam
Period	8/10/16 → 16/10/16

Keywords

CNNs
Cross-depiction problem
Object recognition

ASJC Scopus subject areas

Theoretical Computer Science
Computer Science(all)

Access to Document

10.1007/978-3-319-46604-0_57

Cite this

Westlake, N., Cai, H., & Hall, P. (2016). Detecting people in artwork with CNNs. In G. Hua, & H. Jégou (Eds.), Computer Vision - ECCV 2016 Workshops. ECCV 2016. (pp. 825-841). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 9913 LNCS). Springer Verlag. https://doi.org/10.1007/978-3-319-46604-0_57

Detecting people in artwork with CNNs. / Westlake, Nicholas; Cai, Hongping ; Hall, Peter.

Computer Vision - ECCV 2016 Workshops. ECCV 2016.. ed. / Gang Hua; Hervé Jégou. Springer Verlag, 2016. p. 825-841 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 9913 LNCS).

Research output: Chapter in Book/Report/Conference proceeding › Chapter in a published conference proceeding

Westlake, N, Cai, H & Hall, P 2016, Detecting people in artwork with CNNs. in G Hua & H Jégou (eds), Computer Vision - ECCV 2016 Workshops. ECCV 2016.. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 9913 LNCS, Springer Verlag, pp. 825-841, 14th European Conference on Computer Vision, ECCV 2016, Amsterdam, Netherlands, 8/10/16. https://doi.org/10.1007/978-3-319-46604-0_57

@inproceedings{c50337d1210a48ca8ad9bd3ce29e6ce4,

title = "Detecting people in artwork with CNNs",

abstract = "CNNs have massively improved performance in object detection in photographs. However research into object detection in artwork remains limited. We show state-of-the-art performance on a challenging dataset, People-Art, which contains people from photos, cartoons and 41 different artwork movements. We achieve this high performance by finetuning a CNN for this task, thus also demonstrating that training CNNs on photos results in overfitting for photos: only the first three or four layers transfer from photos to artwork. Although the CNN{\textquoteright}s performance is the highest yet, it remains less than 60% AP, suggesting further work is needed for the cross-depiction problem.",

keywords = "CNNs, Cross-depiction problem, Object recognition",

author = "Nicholas Westlake and Hongping Cai and Peter Hall",

year = "2016",

month = sep,

day = "18",

doi = "10.1007/978-3-319-46604-0_57",

language = "English",

isbn = "9783319466033",

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

publisher = "Springer Verlag",

pages = "825--841",

editor = "Gang Hua and Herv{\'e} J{\'e}gou",

booktitle = "Computer Vision - ECCV 2016 Workshops. ECCV 2016.",

note = "14th European Conference on Computer Vision, ECCV 2016 ; Conference date: 08-10-2016 Through 16-10-2016",

}

TY - GEN

T1 - Detecting people in artwork with CNNs

AU - Westlake, Nicholas

AU - Cai, Hongping

AU - Hall, Peter

PY - 2016/9/18

Y1 - 2016/9/18

N2 - CNNs have massively improved performance in object detection in photographs. However research into object detection in artwork remains limited. We show state-of-the-art performance on a challenging dataset, People-Art, which contains people from photos, cartoons and 41 different artwork movements. We achieve this high performance by finetuning a CNN for this task, thus also demonstrating that training CNNs on photos results in overfitting for photos: only the first three or four layers transfer from photos to artwork. Although the CNN’s performance is the highest yet, it remains less than 60% AP, suggesting further work is needed for the cross-depiction problem.

AB - CNNs have massively improved performance in object detection in photographs. However research into object detection in artwork remains limited. We show state-of-the-art performance on a challenging dataset, People-Art, which contains people from photos, cartoons and 41 different artwork movements. We achieve this high performance by finetuning a CNN for this task, thus also demonstrating that training CNNs on photos results in overfitting for photos: only the first three or four layers transfer from photos to artwork. Although the CNN’s performance is the highest yet, it remains less than 60% AP, suggesting further work is needed for the cross-depiction problem.

KW - CNNs

KW - Cross-depiction problem

KW - Object recognition

UR - http://www.scopus.com/inward/record.url?scp=84989834205&partnerID=8YFLogxK

U2 - 10.1007/978-3-319-46604-0_57

DO - 10.1007/978-3-319-46604-0_57

M3 - Chapter in a published conference proceeding

AN - SCOPUS:84989834205

SN - 9783319466033

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

SP - 825

EP - 841

BT - Computer Vision - ECCV 2016 Workshops. ECCV 2016.

A2 - Hua, Gang

A2 - Jégou, Hervé

PB - Springer Verlag

T2 - 14th European Conference on Computer Vision, ECCV 2016

Y2 - 7 October 2016 through 15 October 2016

ER -

Detecting people in artwork with CNNs

Abstract

Publication series

Conference

Keywords

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Centre for the Analysis of Motion, Entertainment Research and Applications (CAMERA)

Cite this

Detecting people in artwork with CNNs

Abstract

Publication series

Conference

Keywords

ASJC Scopus subject areas

Access to Document

Other files and links

Fingerprint

Projects

Centre for the Analysis of Motion, Entertainment Research and Applications (CAMERA)

Cite this