Improved human parsing with a full relational model

Duan Tran, David Forsyth

Research output: Chapter in Book/Report/Conference proceedingConference 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 languageEnglish (US)
Title of host publicationComputer Vision, ECCV 2010 - 11th European Conference on Computer Vision, Proceedings
PublisherSpringer-Verlag
Pages227-240
Number of pages14
EditionPART 4
ISBN (Print)364215560X, 9783642155604
DOIs
StatePublished - Jan 1 2010
Event11th European Conference on Computer Vision, ECCV 2010 - Heraklion, Crete, Greece
Duration: Sep 10 2010Sep 11 2010

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
NumberPART 4
Volume6314 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Conference

Conference11th European Conference on Computer Vision, ECCV 2010
CountryGreece
CityHeraklion, Crete
Period9/10/109/11/10

Fingerprint

Relational Model
Parsing
Structure Learning
Human
Range of data
Standards

ASJC Scopus subject areas

  • Theoretical Computer Science
  • Computer Science(all)

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 proceedingConference 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
Tran, Duan ; Forsyth, David. / Improved human parsing with a full relational model. Computer Vision, ECCV 2010 - 11th European Conference on Computer Vision, Proceedings. PART 4. ed. Springer-Verlag, 2010. pp. 227-240 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); PART 4).
@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 = "1",
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",

}

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

ER -