SemiStarGAN: Semi-supervised Generative Adversarial Networks for Multi-domain Image-to-Image Translation

Shu Yu Hsu; Chih Yuan Yang; Chi Chia Huang; Jane Yung jen Hsu

doi:10.1007/978-3-030-20870-7_21

SemiStarGAN: Semi-supervised Generative Adversarial Networks for Multi-domain Image-to-Image Translation

Shu Yu Hsu, Chih Yuan Yang, Chi Chia Huang, Jane Yung jen Hsu^*

^*Corresponding author for this work

National Taiwan University

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

3 Scopus citations

Abstract

Recent studies have shown significant advance for multi-domain image-to-image translation, and generative adversarial networks (GANs) are widely used to address this problem. However, to train an effective image generator, existing methods all require a large number of domain-labeled images, which may take time and effort to collect for real-world problems. In this paper, we propose SemiStarGAN, a semi-supervised GAN network to tackle this issue. The proposed method utilizes unlabeled images by incorporating a novel discriminator/classifier network architecture—Y model, and two existing semi-supervised learning techniques—pseudo labeling and self-ensembling. Experimental results on the CelebA dataset using domains of facial attributes show that the proposed method achieves comparable performance with state-of-the-art methods using considerably less labeled training images.

Original language	English
Title of host publication	Computer Vision – ACCV 2018 - 14th Asian Conference on Computer Vision, Revised Selected Papers
Editors	Hongdong Li, Greg Mori, C.V. Jawahar, Konrad Schindler
Publisher	Springer Verlag
Pages	338-353
Number of pages	16
ISBN (Print)	9783030208691
DOIs	https://doi.org/10.1007/978-3-030-20870-7_21
State	Published - 2019
Externally published	Yes
Event	14th Asian Conference on Computer Vision, ACCV 2018 - Perth, Australia Duration: 02 12 2018 → 06 12 2018

Publication series

Name	Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
Volume	11364 LNCS
ISSN (Print)	0302-9743
ISSN (Electronic)	1611-3349

Conference

Conference	14th Asian Conference on Computer Vision, ACCV 2018
Country/Territory	Australia
City	Perth
Period	02/12/18 → 06/12/18

Bibliographical note

Publisher Copyright:
© 2019, Springer Nature Switzerland AG.

Keywords

Generative adversarial network
Image-to-image translation
Semi-supervised learning

Access to Document

10.1007/978-3-030-20870-7_21

Cite this

Hsu, S. Y., Yang, C. Y., Huang, C. C., & Hsu, J. Y. J. (2019). SemiStarGAN: Semi-supervised Generative Adversarial Networks for Multi-domain Image-to-Image Translation. In H. Li, G. Mori, C. V. Jawahar, & K. Schindler (Eds.), Computer Vision – ACCV 2018 - 14th Asian Conference on Computer Vision, Revised Selected Papers (pp. 338-353). (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11364 LNCS). Springer Verlag. https://doi.org/10.1007/978-3-030-20870-7_21

Hsu, Shu Yu ; Yang, Chih Yuan ; Huang, Chi Chia et al. / SemiStarGAN : Semi-supervised Generative Adversarial Networks for Multi-domain Image-to-Image Translation. Computer Vision – ACCV 2018 - 14th Asian Conference on Computer Vision, Revised Selected Papers. editor / Hongdong Li ; Greg Mori ; C.V. Jawahar ; Konrad Schindler. Springer Verlag, 2019. pp. 338-353 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)).

@inproceedings{ced313ca149e440a87d286da6a1e9837,

title = "SemiStarGAN: Semi-supervised Generative Adversarial Networks for Multi-domain Image-to-Image Translation",

abstract = "Recent studies have shown significant advance for multi-domain image-to-image translation, and generative adversarial networks (GANs) are widely used to address this problem. However, to train an effective image generator, existing methods all require a large number of domain-labeled images, which may take time and effort to collect for real-world problems. In this paper, we propose SemiStarGAN, a semi-supervised GAN network to tackle this issue. The proposed method utilizes unlabeled images by incorporating a novel discriminator/classifier network architecture—Y model, and two existing semi-supervised learning techniques—pseudo labeling and self-ensembling. Experimental results on the CelebA dataset using domains of facial attributes show that the proposed method achieves comparable performance with state-of-the-art methods using considerably less labeled training images.",

keywords = "Generative adversarial network, Image-to-image translation, Semi-supervised learning",

author = "Hsu, {Shu Yu} and Yang, {Chih Yuan} and Huang, {Chi Chia} and Hsu, {Jane Yung jen}",

note = "Publisher Copyright: {\textcopyright} 2019, Springer Nature Switzerland AG.; 14th Asian Conference on Computer Vision, ACCV 2018 ; Conference date: 02-12-2018 Through 06-12-2018",

year = "2019",

doi = "10.1007/978-3-030-20870-7_21",

language = "英语",

isbn = "9783030208691",

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

publisher = "Springer Verlag",

pages = "338--353",

editor = "Hongdong Li and Greg Mori and C.V. Jawahar and Konrad Schindler",

booktitle = "Computer Vision – ACCV 2018 - 14th Asian Conference on Computer Vision, Revised Selected Papers",

address = "德国",

}

Hsu, SY , Yang, CY , Huang, CC & Hsu, JYJ 2019, SemiStarGAN: Semi-supervised Generative Adversarial Networks for Multi-domain Image-to-Image Translation. in H Li, G Mori, CV Jawahar & K Schindler (eds), Computer Vision – ACCV 2018 - 14th Asian Conference on Computer Vision, Revised Selected Papers. Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics), vol. 11364 LNCS, Springer Verlag, pp. 338-353, 14th Asian Conference on Computer Vision, ACCV 2018, Perth, Australia, 02/12/18. https://doi.org/10.1007/978-3-030-20870-7_21

SemiStarGAN: Semi-supervised Generative Adversarial Networks for Multi-domain Image-to-Image Translation. / Hsu, Shu Yu ; Yang, Chih Yuan ; Huang, Chi Chia et al.
Computer Vision – ACCV 2018 - 14th Asian Conference on Computer Vision, Revised Selected Papers. ed. / Hongdong Li; Greg Mori; C.V. Jawahar; Konrad Schindler. Springer Verlag, 2019. p. 338-353 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 11364 LNCS).

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

TY - GEN

T1 - SemiStarGAN

T2 - 14th Asian Conference on Computer Vision, ACCV 2018

AU - Hsu, Shu Yu

AU - Yang, Chih Yuan

AU - Huang, Chi Chia

AU - Hsu, Jane Yung jen

PY - 2019

Y1 - 2019

N2 - Recent studies have shown significant advance for multi-domain image-to-image translation, and generative adversarial networks (GANs) are widely used to address this problem. However, to train an effective image generator, existing methods all require a large number of domain-labeled images, which may take time and effort to collect for real-world problems. In this paper, we propose SemiStarGAN, a semi-supervised GAN network to tackle this issue. The proposed method utilizes unlabeled images by incorporating a novel discriminator/classifier network architecture—Y model, and two existing semi-supervised learning techniques—pseudo labeling and self-ensembling. Experimental results on the CelebA dataset using domains of facial attributes show that the proposed method achieves comparable performance with state-of-the-art methods using considerably less labeled training images.

AB - Recent studies have shown significant advance for multi-domain image-to-image translation, and generative adversarial networks (GANs) are widely used to address this problem. However, to train an effective image generator, existing methods all require a large number of domain-labeled images, which may take time and effort to collect for real-world problems. In this paper, we propose SemiStarGAN, a semi-supervised GAN network to tackle this issue. The proposed method utilizes unlabeled images by incorporating a novel discriminator/classifier network architecture—Y model, and two existing semi-supervised learning techniques—pseudo labeling and self-ensembling. Experimental results on the CelebA dataset using domains of facial attributes show that the proposed method achieves comparable performance with state-of-the-art methods using considerably less labeled training images.

KW - Generative adversarial network

KW - Image-to-image translation

KW - Semi-supervised learning

UR - http://www.scopus.com/inward/record.url?scp=85066872238&partnerID=8YFLogxK

U2 - 10.1007/978-3-030-20870-7_21

DO - 10.1007/978-3-030-20870-7_21

M3 - 会议稿件

AN - SCOPUS:85066872238

SN - 9783030208691

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

SP - 338

EP - 353

BT - Computer Vision – ACCV 2018 - 14th Asian Conference on Computer Vision, Revised Selected Papers

A2 - Li, Hongdong

A2 - Mori, Greg

A2 - Jawahar, C.V.

A2 - Schindler, Konrad

PB - Springer Verlag

Y2 - 2 December 2018 through 6 December 2018

ER -

Hsu SY , Yang CY , Huang CC, Hsu JYJ. SemiStarGAN: Semi-supervised Generative Adversarial Networks for Multi-domain Image-to-Image Translation. In Li H, Mori G, Jawahar CV, Schindler K, editors, Computer Vision – ACCV 2018 - 14th Asian Conference on Computer Vision, Revised Selected Papers. Springer Verlag. 2019. p. 338-353. (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)). doi: 10.1007/978-3-030-20870-7_21

SemiStarGAN: Semi-supervised Generative Adversarial Networks for Multi-domain Image-to-Image Translation

Abstract

Publication series

Conference

Bibliographical note

Keywords

Access to Document

Other files and links

Fingerprint

Cite this