Improved human parsing with a full relational model

Duan Tran; David Forsyth

doi:10.1007/978-3-642-15561-1_17

Improved human parsing with a full relational model

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

Abstract

We show quantitative evidence that a full relational model of the body performs better at upper body parsing than the standard tree model, despite the need to adopt approximate inference and learning procedures. Our method uses an approximate search for inference, and an approximate structure learning method to learn. We compare our method to state of the art methods on our dataset (which depicts a wide range of poses), on the standard Buffy dataset, and on the reduced PASCAL dataset published recently. Our results suggest that the Buffy dataset over emphasizes poses where the arms hang down, and that leads to generalization problems.

Original language	English (US)
Title of host publication	Computer Vision, ECCV 2010 - 11th European Conference on Computer Vision, Proceedings
Publisher	Springer-Verlag
Pages	227-240
Number of pages	14
Edition	PART 4
ISBN (Print)	364215560X, 9783642155604
DOIs	https://doi.org/10.1007/978-3-642-15561-1_17
State	Published - Jan 1 2010
Event	11th European Conference on Computer Vision, ECCV 2010 - Heraklion, Crete, Greece Duration: Sep 10 2010 → Sep 11 2010

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Number	PART 4
Volume	6314 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	11th European Conference on Computer Vision, ECCV 2010
Country	Greece
City	Heraklion, Crete
Period	9/10/10 → 9/11/10

ASJC Scopus subject areas

Theoretical Computer Science
Computer Science(all)

Access to Document

10.1007/978-3-642-15561-1_17

Fingerprint Dive into the research topics of 'Improved human parsing with a full relational model'. Together they form a unique fingerprint.

Cite this

Tran, D., & Forsyth, D. (2010). Improved human parsing with a full relational model. In Computer Vision, ECCV 2010 - 11th European Conference on Computer Vision, Proceedings (PART 4 ed., pp. 227-240). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 6314 LNCS, No. PART 4). Springer-Verlag. https://doi.org/10.1007/978-3-642-15561-1_17

Improved human parsing with a full relational model. / Tran, Duan; Forsyth, David.

Computer Vision, ECCV 2010 - 11th European Conference on Computer Vision, Proceedings. PART 4. ed. Springer-Verlag, 2010. p. 227-240 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 6314 LNCS, No. PART 4).

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

Tran, D & Forsyth, D 2010, Improved human parsing with a full relational model. in Computer Vision, ECCV 2010 - 11th European Conference on Computer Vision, Proceedings. PART 4 edn, Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), no. PART 4, vol. 6314 LNCS, Springer-Verlag, pp. 227-240, 11th European Conference on Computer Vision, ECCV 2010, Heraklion, Crete, Greece, 9/10/10. https://doi.org/10.1007/978-3-642-15561-1_17

Tran D, Forsyth D. Improved human parsing with a full relational model. In Computer Vision, ECCV 2010 - 11th European Conference on Computer Vision, Proceedings. PART 4 ed. Springer-Verlag. 2010. p. 227-240. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); PART 4). https://doi.org/10.1007/978-3-642-15561-1_17

@inproceedings{c200df14d52f4f5f9e5538283a210f20,

title = "Improved human parsing with a full relational model",

abstract = "We show quantitative evidence that a full relational model of the body performs better at upper body parsing than the standard tree model, despite the need to adopt approximate inference and learning procedures. Our method uses an approximate search for inference, and an approximate structure learning method to learn. We compare our method to state of the art methods on our dataset (which depicts a wide range of poses), on the standard Buffy dataset, and on the reduced PASCAL dataset published recently. Our results suggest that the Buffy dataset over emphasizes poses where the arms hang down, and that leads to generalization problems.",

author = "Duan Tran and David Forsyth",

year = "2010",

month = jan,

day = "1",

doi = "10.1007/978-3-642-15561-1_17",

language = "English (US)",

isbn = "364215560X",

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

publisher = "Springer-Verlag",

number = "PART 4",

pages = "227--240",

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

edition = "PART 4",

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

}

TY - GEN

T1 - Improved human parsing with a full relational model

AU - Tran, Duan

AU - Forsyth, David

PY - 2010/1/1

Y1 - 2010/1/1

N2 - We show quantitative evidence that a full relational model of the body performs better at upper body parsing than the standard tree model, despite the need to adopt approximate inference and learning procedures. Our method uses an approximate search for inference, and an approximate structure learning method to learn. We compare our method to state of the art methods on our dataset (which depicts a wide range of poses), on the standard Buffy dataset, and on the reduced PASCAL dataset published recently. Our results suggest that the Buffy dataset over emphasizes poses where the arms hang down, and that leads to generalization problems.

AB - We show quantitative evidence that a full relational model of the body performs better at upper body parsing than the standard tree model, despite the need to adopt approximate inference and learning procedures. Our method uses an approximate search for inference, and an approximate structure learning method to learn. We compare our method to state of the art methods on our dataset (which depicts a wide range of poses), on the standard Buffy dataset, and on the reduced PASCAL dataset published recently. Our results suggest that the Buffy dataset over emphasizes poses where the arms hang down, and that leads to generalization problems.

UR - http://www.scopus.com/inward/record.url?scp=78149316204&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=78149316204&partnerID=8YFLogxK

U2 - 10.1007/978-3-642-15561-1_17

DO - 10.1007/978-3-642-15561-1_17

M3 - Conference contribution

AN - SCOPUS:78149316204

SN - 364215560X

SN - 9783642155604

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

SP - 227

EP - 240

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 -