Regularizing Generative Adversarial Networks under Limited Data

Hung Yu Tseng; Lu Jiang; Ce Liu; Ming Hsuan Yang; Weilong Yang

doi:10.1109/CVPR46437.2021.00783

Regularizing Generative Adversarial Networks under Limited Data

Hung Yu Tseng, Lu Jiang, Ce Liu, Ming Hsuan Yang, Weilong Yang

College of Computing

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

25 Citations (Scopus)

Abstract

Recent years have witnessed the rapid progress of generative adversarial networks (GANs). However, the success of the GAN models hinges on a large amount of training data. This work proposes a regularization approach for training robust GAN models on limited data. We theoretically show a connection between the regularized loss and an f-divergence called LeCam-divergence, which we find is more robust under limited training data. Extensive experiments on several benchmark datasets demonstrate that the proposed regularization scheme 1) improves the generalization performance and stabilizes the learning dynamics of GAN models under limited training data, and 2) complements the recent data augmentation methods. These properties facilitate training GAN models to achieve state-of-the-art performance when only limited training data of the ImageNet benchmark is available. The source code is available at https://github.com/google/lecam-gan.

Original language	English
Title of host publication	Proceedings - 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2021
Publisher	IEEE Computer Society
Pages	7917-7927
Number of pages	11
ISBN (Electronic)	9781665445092
DOIs	https://doi.org/10.1109/CVPR46437.2021.00783
Publication status	Published - 2021
Event	2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2021 - Virtual, Online, United States Duration: 2021 Jun 19 → 2021 Jun 25

Publication series

Name	Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition
ISSN (Print)	1063-6919

Conference

Conference	2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2021
Country/Territory	United States
City	Virtual, Online
Period	21/6/19 → 21/6/25

Bibliographical note

Funding Information:
This work is supported in part by the NSF CAREER Grant #1149783.

Publisher Copyright:
© 2021 IEEE

All Science Journal Classification (ASJC) codes

Software
Computer Vision and Pattern Recognition

Access to Document

10.1109/CVPR46437.2021.00783

Cite this

Tseng, H. Y., Jiang, L., Liu, C., Yang, M. H., & Yang, W. (2021). Regularizing Generative Adversarial Networks under Limited Data. In Proceedings - 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2021 (pp. 7917-7927). (Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition). IEEE Computer Society. https://doi.org/10.1109/CVPR46437.2021.00783

@inproceedings{86b2fa6d9a9c450c860e489e208ec9eb,

title = "Regularizing Generative Adversarial Networks under Limited Data",

abstract = "Recent years have witnessed the rapid progress of generative adversarial networks (GANs). However, the success of the GAN models hinges on a large amount of training data. This work proposes a regularization approach for training robust GAN models on limited data. We theoretically show a connection between the regularized loss and an f-divergence called LeCam-divergence, which we find is more robust under limited training data. Extensive experiments on several benchmark datasets demonstrate that the proposed regularization scheme 1) improves the generalization performance and stabilizes the learning dynamics of GAN models under limited training data, and 2) complements the recent data augmentation methods. These properties facilitate training GAN models to achieve state-of-the-art performance when only limited training data of the ImageNet benchmark is available. The source code is available at https://github.com/google/lecam-gan.",

author = "Tseng, {Hung Yu} and Lu Jiang and Ce Liu and Yang, {Ming Hsuan} and Weilong Yang",

note = "Funding Information: This work is supported in part by the NSF CAREER Grant #1149783. Publisher Copyright: {\textcopyright} 2021 IEEE; 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2021 ; Conference date: 19-06-2021 Through 25-06-2021",

year = "2021",

doi = "10.1109/CVPR46437.2021.00783",

language = "English",

series = "Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition",

publisher = "IEEE Computer Society",

pages = "7917--7927",

booktitle = "Proceedings - 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2021",

address = "United States",

}

Tseng, HY, Jiang, L, Liu, C, Yang, MH & Yang, W 2021, Regularizing Generative Adversarial Networks under Limited Data. in Proceedings - 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2021. Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition, IEEE Computer Society, pp. 7917-7927, 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2021, Virtual, Online, United States, 21/6/19. https://doi.org/10.1109/CVPR46437.2021.00783

Regularizing Generative Adversarial Networks under Limited Data. / Tseng, Hung Yu; Jiang, Lu; Liu, Ce et al.

Proceedings - 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2021. IEEE Computer Society, 2021. p. 7917-7927 (Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition).

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

TY - GEN

T1 - Regularizing Generative Adversarial Networks under Limited Data

AU - Tseng, Hung Yu

AU - Jiang, Lu

AU - Liu, Ce

AU - Yang, Ming Hsuan

AU - Yang, Weilong

PY - 2021

Y1 - 2021

N2 - Recent years have witnessed the rapid progress of generative adversarial networks (GANs). However, the success of the GAN models hinges on a large amount of training data. This work proposes a regularization approach for training robust GAN models on limited data. We theoretically show a connection between the regularized loss and an f-divergence called LeCam-divergence, which we find is more robust under limited training data. Extensive experiments on several benchmark datasets demonstrate that the proposed regularization scheme 1) improves the generalization performance and stabilizes the learning dynamics of GAN models under limited training data, and 2) complements the recent data augmentation methods. These properties facilitate training GAN models to achieve state-of-the-art performance when only limited training data of the ImageNet benchmark is available. The source code is available at https://github.com/google/lecam-gan.

AB - Recent years have witnessed the rapid progress of generative adversarial networks (GANs). However, the success of the GAN models hinges on a large amount of training data. This work proposes a regularization approach for training robust GAN models on limited data. We theoretically show a connection between the regularized loss and an f-divergence called LeCam-divergence, which we find is more robust under limited training data. Extensive experiments on several benchmark datasets demonstrate that the proposed regularization scheme 1) improves the generalization performance and stabilizes the learning dynamics of GAN models under limited training data, and 2) complements the recent data augmentation methods. These properties facilitate training GAN models to achieve state-of-the-art performance when only limited training data of the ImageNet benchmark is available. The source code is available at https://github.com/google/lecam-gan.

UR - http://www.scopus.com/inward/record.url?scp=85122360763&partnerID=8YFLogxK

UR - http://www.scopus.com/inward/citedby.url?scp=85122360763&partnerID=8YFLogxK

U2 - 10.1109/CVPR46437.2021.00783

DO - 10.1109/CVPR46437.2021.00783

M3 - Conference contribution

AN - SCOPUS:85122360763

T3 - Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition

SP - 7917

EP - 7927

BT - Proceedings - 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2021

PB - IEEE Computer Society

T2 - 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2021

Y2 - 19 June 2021 through 25 June 2021

ER -

Tseng HY, Jiang L, Liu C, Yang MH, Yang W. Regularizing Generative Adversarial Networks under Limited Data. In Proceedings - 2021 IEEE/CVF Conference on Computer Vision and Pattern Recognition, CVPR 2021. IEEE Computer Society. 2021. p. 7917-7927. (Proceedings of the IEEE Computer Society Conference on Computer Vision and Pattern Recognition). doi: 10.1109/CVPR46437.2021.00783

Regularizing Generative Adversarial Networks under Limited Data

Abstract

Publication series

Conference

Bibliographical note

All Science Journal Classification (ASJC) codes

Access to Document

Other files and links

Fingerprint

Cite this