Efficient relative attribute learning using graph neural networks

Zihang Meng, Nagesh Adluru, Hyun Woo Kim, Glenn Fung, Vikas Singh

Research output: Chapter in Book/Report/Conference proceedingConference contribution

1 Citation (Scopus)

Abstract

A sizable body of work on relative attributes provides evidence that relating pairs of images along a continuum of strength pertaining to a visual attribute yields improvements in a variety of vision tasks. In this paper, we show how emerging ideas in graph neural networks can yield a solution to various problems that broadly fall under relative attribute learning. Our main idea is the observation that relative attribute learning naturally benefits from exploiting the graph of dependencies among the different relative attributes of images, especially when only partial ordering is provided at training time. We use message passing to perform end to end learning of the image representations, their relationships as well as the interplay between different attributes. Our experiments show that this simple framework is effective in achieving competitive accuracy with specialized methods for both relative attribute learning and binary attribute prediction, while relaxing the requirements on the training data and/or the number of parameters, or both.

Original languageEnglish
Title of host publicationComputer Vision – ECCV 2018 - 15th European Conference, 2018, Proceedings
EditorsVittorio Ferrari, Cristian Sminchisescu, Yair Weiss, Martial Hebert
PublisherSpringer Verlag
Pages575-590
Number of pages16
ISBN (Print)9783030012632
DOIs
Publication statusPublished - 2018 Jan 1
Externally publishedYes
Event15th European Conference on Computer Vision, ECCV 2018 - Munich, Germany
Duration: 2018 Sep 82018 Sep 14

Publication series

NameLecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume11218 LNCS
ISSN (Print)0302-9743
ISSN (Electronic)1611-3349

Other

Other15th European Conference on Computer Vision, ECCV 2018
CountryGermany
CityMunich
Period18/9/818/9/14

Fingerprint

Message passing
Attribute
Neural Networks
Neural networks
Graph in graph theory
Experiments
Image Representation
Partial ordering
Learning
Message Passing
Continuum
Binary
Prediction
Requirements
Experiment

Keywords

  • Graph neural networks
  • Message passing
  • Multi-task learning
  • Relative attribute learning

ASJC Scopus subject areas

  • Theoretical Computer Science
  • Computer Science(all)

Cite this

Meng, Z., Adluru, N., Kim, H. W., Fung, G., & Singh, V. (2018). Efficient relative attribute learning using graph neural networks. In V. Ferrari, C. Sminchisescu, Y. Weiss, & M. Hebert (Eds.), Computer Vision – ECCV 2018 - 15th European Conference, 2018, Proceedings (pp. 575-590). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11218 LNCS). Springer Verlag. https://doi.org/10.1007/978-3-030-01264-9_34

Efficient relative attribute learning using graph neural networks. / Meng, Zihang; Adluru, Nagesh; Kim, Hyun Woo; Fung, Glenn; Singh, Vikas.

Computer Vision – ECCV 2018 - 15th European Conference, 2018, Proceedings. ed. / Vittorio Ferrari; Cristian Sminchisescu; Yair Weiss; Martial Hebert. Springer Verlag, 2018. p. 575-590 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11218 LNCS).

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Meng, Z, Adluru, N, Kim, HW, Fung, G & Singh, V 2018, Efficient relative attribute learning using graph neural networks. in V Ferrari, C Sminchisescu, Y Weiss & M Hebert (eds), Computer Vision – ECCV 2018 - 15th European Conference, 2018, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 11218 LNCS, Springer Verlag, pp. 575-590, 15th European Conference on Computer Vision, ECCV 2018, Munich, Germany, 18/9/8. https://doi.org/10.1007/978-3-030-01264-9_34
Meng Z, Adluru N, Kim HW, Fung G, Singh V. Efficient relative attribute learning using graph neural networks. In Ferrari V, Sminchisescu C, Weiss Y, Hebert M, editors, Computer Vision – ECCV 2018 - 15th European Conference, 2018, Proceedings. Springer Verlag. 2018. p. 575-590. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). https://doi.org/10.1007/978-3-030-01264-9_34
Meng, Zihang ; Adluru, Nagesh ; Kim, Hyun Woo ; Fung, Glenn ; Singh, Vikas. / Efficient relative attribute learning using graph neural networks. Computer Vision – ECCV 2018 - 15th European Conference, 2018, Proceedings. editor / Vittorio Ferrari ; Cristian Sminchisescu ; Yair Weiss ; Martial Hebert. Springer Verlag, 2018. pp. 575-590 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).
@inproceedings{54dc84f5b5ea41b9b64e49900690a21b,
title = "Efficient relative attribute learning using graph neural networks",
abstract = "A sizable body of work on relative attributes provides evidence that relating pairs of images along a continuum of strength pertaining to a visual attribute yields improvements in a variety of vision tasks. In this paper, we show how emerging ideas in graph neural networks can yield a solution to various problems that broadly fall under relative attribute learning. Our main idea is the observation that relative attribute learning naturally benefits from exploiting the graph of dependencies among the different relative attributes of images, especially when only partial ordering is provided at training time. We use message passing to perform end to end learning of the image representations, their relationships as well as the interplay between different attributes. Our experiments show that this simple framework is effective in achieving competitive accuracy with specialized methods for both relative attribute learning and binary attribute prediction, while relaxing the requirements on the training data and/or the number of parameters, or both.",
keywords = "Graph neural networks, Message passing, Multi-task learning, Relative attribute learning",
author = "Zihang Meng and Nagesh Adluru and Kim, {Hyun Woo} and Glenn Fung and Vikas Singh",
year = "2018",
month = "1",
day = "1",
doi = "10.1007/978-3-030-01264-9_34",
language = "English",
isbn = "9783030012632",
series = "Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)",
publisher = "Springer Verlag",
pages = "575--590",
editor = "Vittorio Ferrari and Cristian Sminchisescu and Yair Weiss and Martial Hebert",
booktitle = "Computer Vision – ECCV 2018 - 15th European Conference, 2018, Proceedings",

}

TY - GEN

T1 - Efficient relative attribute learning using graph neural networks

AU - Meng, Zihang

AU - Adluru, Nagesh

AU - Kim, Hyun Woo

AU - Fung, Glenn

AU - Singh, Vikas

PY - 2018/1/1

Y1 - 2018/1/1

N2 - A sizable body of work on relative attributes provides evidence that relating pairs of images along a continuum of strength pertaining to a visual attribute yields improvements in a variety of vision tasks. In this paper, we show how emerging ideas in graph neural networks can yield a solution to various problems that broadly fall under relative attribute learning. Our main idea is the observation that relative attribute learning naturally benefits from exploiting the graph of dependencies among the different relative attributes of images, especially when only partial ordering is provided at training time. We use message passing to perform end to end learning of the image representations, their relationships as well as the interplay between different attributes. Our experiments show that this simple framework is effective in achieving competitive accuracy with specialized methods for both relative attribute learning and binary attribute prediction, while relaxing the requirements on the training data and/or the number of parameters, or both.

AB - A sizable body of work on relative attributes provides evidence that relating pairs of images along a continuum of strength pertaining to a visual attribute yields improvements in a variety of vision tasks. In this paper, we show how emerging ideas in graph neural networks can yield a solution to various problems that broadly fall under relative attribute learning. Our main idea is the observation that relative attribute learning naturally benefits from exploiting the graph of dependencies among the different relative attributes of images, especially when only partial ordering is provided at training time. We use message passing to perform end to end learning of the image representations, their relationships as well as the interplay between different attributes. Our experiments show that this simple framework is effective in achieving competitive accuracy with specialized methods for both relative attribute learning and binary attribute prediction, while relaxing the requirements on the training data and/or the number of parameters, or both.

KW - Graph neural networks

KW - Message passing

KW - Multi-task learning

KW - Relative attribute learning

UR - http://www.scopus.com/inward/record.url?scp=85055714853&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85055714853&partnerID=8YFLogxK

U2 - 10.1007/978-3-030-01264-9_34

DO - 10.1007/978-3-030-01264-9_34

M3 - Conference contribution

AN - SCOPUS:85055714853

SN - 9783030012632

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

SP - 575

EP - 590

BT - Computer Vision – ECCV 2018 - 15th European Conference, 2018, Proceedings

A2 - Ferrari, Vittorio

A2 - Sminchisescu, Cristian

A2 - Weiss, Yair

A2 - Hebert, Martial

PB - Springer Verlag

ER -