Multi-label Generalized Zero-Shot Learning Using Identifiable Variational Autoencoders

Muqaddas Gull; Omar Arif

doi:10.1007/978-3-031-43404-4_3

Multi-label Generalized Zero-Shot Learning Using Identifiable Variational Autoencoders

Muqaddas Gull^*, Omar Arif

^*Corresponding author for this work

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

2 Scopus citations

Abstract

Multi-label Zero-Shot Learning (ZSL) is an extension of traditional single-label ZSL, where the objective is to accurately classify images containing multiple unseen classes that are not available during training. Current techniques depends on attention mechanisms and Generative Adversarial Networks (GAN) to address multi-label ZSL and Generalized Zero-Shot Learning (GZSL) challenge. However, generating features for both multi-label ZSL and GZSL in the context of disentangled representation learning remains unexplored. In this paper, we propose an identifiable Variational Autoencoder (iVAE) based generative framework for multi-label ZSL and GZSL. The main idea of our proposed approach is to learn disentangled representations for generating semantically consistent multi-label features using an attribute-level feature fusion technique. We perform comprehensive experiments on two benchmark datasets, NUS-WIDE and MS COCO, for both multi-label ZSL and GZSL. Furthermore, disentangled representation learning for both multi-label ZSL and GZSL on standard datasets achieves commendable performance as compared to existing methods.

Original language	English
Title of host publication	Extended Reality - International Conference, XR Salento 2023, Proceedings
Editors	Lucio Tommaso De Paolis, Pasquale Arpaia, Marco Sacco
Publisher	Springer Science and Business Media Deutschland GmbH
Pages	35-50
Number of pages	16
ISBN (Print)	9783031434037
DOIs	https://doi.org/10.1007/978-3-031-43404-4_3
State	Published - 2023
Externally published	Yes
Event	Proceedings of the International Conference on extended Reality, XR SALENTO 2023 - Lecce, Italy Duration: 6 Sep 2023 → 9 Sep 2023

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	14219 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	Proceedings of the International Conference on extended Reality, XR SALENTO 2023
Country/Territory	Italy
City	Lecce
Period	6/09/23 → 9/09/23

Bibliographical note

Publisher Copyright:
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2023.

Keywords

Attribute-Level Feature Fusion
Disentangled Representation Learning
Generalized Zero-Shot Learning
Zero-Shot Learning

ASJC Scopus subject areas

Theoretical Computer Science
General Computer Science

Access to Document

10.1007/978-3-031-43404-4_3

Cite this

Gull, M., & Arif, O. (2023). Multi-label Generalized Zero-Shot Learning Using Identifiable Variational Autoencoders. In L. T. De Paolis, P. Arpaia, & M. Sacco (Eds.), Extended Reality - International Conference, XR Salento 2023, Proceedings (pp. 35-50). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 14219 LNCS). Springer Science and Business Media Deutschland GmbH. https://doi.org/10.1007/978-3-031-43404-4_3

Gull, Muqaddas ; Arif, Omar. / Multi-label Generalized Zero-Shot Learning Using Identifiable Variational Autoencoders. Extended Reality - International Conference, XR Salento 2023, Proceedings. editor / Lucio Tommaso De Paolis ; Pasquale Arpaia ; Marco Sacco. Springer Science and Business Media Deutschland GmbH, 2023. pp. 35-50 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{93400ddb13774813bcb8460365d9bcc6,

title = "Multi-label Generalized Zero-Shot Learning Using Identifiable Variational Autoencoders",

abstract = "Multi-label Zero-Shot Learning (ZSL) is an extension of traditional single-label ZSL, where the objective is to accurately classify images containing multiple unseen classes that are not available during training. Current techniques depends on attention mechanisms and Generative Adversarial Networks (GAN) to address multi-label ZSL and Generalized Zero-Shot Learning (GZSL) challenge. However, generating features for both multi-label ZSL and GZSL in the context of disentangled representation learning remains unexplored. In this paper, we propose an identifiable Variational Autoencoder (iVAE) based generative framework for multi-label ZSL and GZSL. The main idea of our proposed approach is to learn disentangled representations for generating semantically consistent multi-label features using an attribute-level feature fusion technique. We perform comprehensive experiments on two benchmark datasets, NUS-WIDE and MS COCO, for both multi-label ZSL and GZSL. Furthermore, disentangled representation learning for both multi-label ZSL and GZSL on standard datasets achieves commendable performance as compared to existing methods.",

keywords = "Attribute-Level Feature Fusion, Disentangled Representation Learning, Generalized Zero-Shot Learning, Zero-Shot Learning",

author = "Muqaddas Gull and Omar Arif",

note = "Publisher Copyright: {\textcopyright} The Author(s), under exclusive license to Springer Nature Switzerland AG 2023.; Proceedings of the International Conference on extended Reality, XR SALENTO 2023 ; Conference date: 06-09-2023 Through 09-09-2023",

year = "2023",

doi = "10.1007/978-3-031-43404-4\_3",

language = "English",

isbn = "9783031434037",

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 = "35--50",

editor = "\{De Paolis\}, \{Lucio Tommaso\} and Pasquale Arpaia and Marco Sacco",

booktitle = "Extended Reality - International Conference, XR Salento 2023, Proceedings",

address = "Germany",

}

Gull, M & Arif, O 2023, Multi-label Generalized Zero-Shot Learning Using Identifiable Variational Autoencoders. in LT De Paolis, P Arpaia & M Sacco (eds), Extended Reality - International Conference, XR Salento 2023, Proceedings. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 14219 LNCS, Springer Science and Business Media Deutschland GmbH, pp. 35-50, Proceedings of the International Conference on extended Reality, XR SALENTO 2023, Lecce, Italy, 6/09/23. https://doi.org/10.1007/978-3-031-43404-4_3

Multi-label Generalized Zero-Shot Learning Using Identifiable Variational Autoencoders. / Gull, Muqaddas; Arif, Omar.
Extended Reality - International Conference, XR Salento 2023, Proceedings. ed. / Lucio Tommaso De Paolis; Pasquale Arpaia; Marco Sacco. Springer Science and Business Media Deutschland GmbH, 2023. p. 35-50 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 14219 LNCS).

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

TY - GEN

T1 - Multi-label Generalized Zero-Shot Learning Using Identifiable Variational Autoencoders

AU - Gull, Muqaddas

AU - Arif, Omar

N1 - Publisher Copyright: © The Author(s), under exclusive license to Springer Nature Switzerland AG 2023.

PY - 2023

Y1 - 2023

N2 - Multi-label Zero-Shot Learning (ZSL) is an extension of traditional single-label ZSL, where the objective is to accurately classify images containing multiple unseen classes that are not available during training. Current techniques depends on attention mechanisms and Generative Adversarial Networks (GAN) to address multi-label ZSL and Generalized Zero-Shot Learning (GZSL) challenge. However, generating features for both multi-label ZSL and GZSL in the context of disentangled representation learning remains unexplored. In this paper, we propose an identifiable Variational Autoencoder (iVAE) based generative framework for multi-label ZSL and GZSL. The main idea of our proposed approach is to learn disentangled representations for generating semantically consistent multi-label features using an attribute-level feature fusion technique. We perform comprehensive experiments on two benchmark datasets, NUS-WIDE and MS COCO, for both multi-label ZSL and GZSL. Furthermore, disentangled representation learning for both multi-label ZSL and GZSL on standard datasets achieves commendable performance as compared to existing methods.

AB - Multi-label Zero-Shot Learning (ZSL) is an extension of traditional single-label ZSL, where the objective is to accurately classify images containing multiple unseen classes that are not available during training. Current techniques depends on attention mechanisms and Generative Adversarial Networks (GAN) to address multi-label ZSL and Generalized Zero-Shot Learning (GZSL) challenge. However, generating features for both multi-label ZSL and GZSL in the context of disentangled representation learning remains unexplored. In this paper, we propose an identifiable Variational Autoencoder (iVAE) based generative framework for multi-label ZSL and GZSL. The main idea of our proposed approach is to learn disentangled representations for generating semantically consistent multi-label features using an attribute-level feature fusion technique. We perform comprehensive experiments on two benchmark datasets, NUS-WIDE and MS COCO, for both multi-label ZSL and GZSL. Furthermore, disentangled representation learning for both multi-label ZSL and GZSL on standard datasets achieves commendable performance as compared to existing methods.

KW - Attribute-Level Feature Fusion

KW - Disentangled Representation Learning

KW - Generalized Zero-Shot Learning

KW - Zero-Shot Learning

UR - https://www.scopus.com/pages/publications/85173588820

U2 - 10.1007/978-3-031-43404-4_3

DO - 10.1007/978-3-031-43404-4_3

M3 - Conference contribution

AN - SCOPUS:85173588820

SN - 9783031434037

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

SP - 35

EP - 50

BT - Extended Reality - International Conference, XR Salento 2023, Proceedings

A2 - De Paolis, Lucio Tommaso

A2 - Arpaia, Pasquale

A2 - Sacco, Marco

PB - Springer Science and Business Media Deutschland GmbH

T2 - Proceedings of the International Conference on extended Reality, XR SALENTO 2023

Y2 - 6 September 2023 through 9 September 2023

ER -

Gull M, Arif O. Multi-label Generalized Zero-Shot Learning Using Identifiable Variational Autoencoders. In De Paolis LT, Arpaia P, Sacco M, editors, Extended Reality - International Conference, XR Salento 2023, Proceedings. Springer Science and Business Media Deutschland GmbH. 2023. p. 35-50. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-031-43404-4_3

Multi-label Generalized Zero-Shot Learning Using Identifiable Variational Autoencoders

Abstract

Publication series

Conference

Bibliographical note

Keywords

ASJC Scopus subject areas

Access to Document

Fingerprint

Cite this