State-of-the-art methods for image-to-image translation with Generative Adversarial Networks (GANs) can learn a mapping from one domain to another domain using unpaired image data. However, these methods require the training of one specific model for every pair of image domains, which limits the scalability in dealing with more than two image domains. In addition, the training stage of these methods has the common problem of model collapse that degrades the quality of the generated images. To tackle these issues, we propose a Dual Generator Generative Adversarial Network (G2GAN), which is a robust and scalable approach allowing to perform unpaired image-to-image translation for multiple domains using only dual generators within a single model. Moreover, we explore different optimization losses for better training of G2GAN, and thus make unpaired image-to-image translation with higher consistency and better stability. Extensive experiments on six publicly available datasets with different scenarios, i.e., architectural buildings, seasons, landscape and human faces, demonstrate that the proposed G2GAN achieves superior model capacity and better generation performance comparing with existing image-to-image translation GAN models.

Dual Generator Generative Adversarial Networks for Multi-domain Image-to-Image Translation / Tang, H.; Xu, D.; Wang, W.; Yan, Yan; Sebe, Niculae. - 11361 LNCS:(2019), pp. 3-21. (Intervento presentato al convegno ACCV tenutosi a Perth nel 2–6 December, 2018) [10.1007/978-3-030-20887-5_1].

Dual Generator Generative Adversarial Networks for Multi-domain Image-to-Image Translation

H. Tang;D. Xu;W. Wang;Y. Yan;N. Sebe.
2019-01-01

Abstract

State-of-the-art methods for image-to-image translation with Generative Adversarial Networks (GANs) can learn a mapping from one domain to another domain using unpaired image data. However, these methods require the training of one specific model for every pair of image domains, which limits the scalability in dealing with more than two image domains. In addition, the training stage of these methods has the common problem of model collapse that degrades the quality of the generated images. To tackle these issues, we propose a Dual Generator Generative Adversarial Network (G2GAN), which is a robust and scalable approach allowing to perform unpaired image-to-image translation for multiple domains using only dual generators within a single model. Moreover, we explore different optimization losses for better training of G2GAN, and thus make unpaired image-to-image translation with higher consistency and better stability. Extensive experiments on six publicly available datasets with different scenarios, i.e., architectural buildings, seasons, landscape and human faces, demonstrate that the proposed G2GAN achieves superior model capacity and better generation performance comparing with existing image-to-image translation GAN models.
2019
Asian Conference on Computer Vision 2018
Heidelberg
Springer
Tang, H.; Xu, D.; Wang, W.; Yan, Yan; Sebe, Niculae
Dual Generator Generative Adversarial Networks for Multi-domain Image-to-Image Translation / Tang, H.; Xu, D.; Wang, W.; Yan, Yan; Sebe, Niculae. - 11361 LNCS:(2019), pp. 3-21. (Intervento presentato al convegno ACCV tenutosi a Perth nel 2–6 December, 2018) [10.1007/978-3-030-20887-5_1].
File in questo prodotto:
File Dimensione Formato  
1901.04604.pdf

accesso aperto

Tipologia: Post-print referato (Refereed author’s manuscript)
Licenza: Tutti i diritti riservati (All rights reserved)
Dimensione 8.9 MB
Formato Adobe PDF
8.9 MB Adobe PDF Visualizza/Apri

I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione

Utilizza questo identificativo per citare o creare un link a questo documento: https://hdl.handle.net/11572/247610
Citazioni
  • ???jsp.display-item.citation.pmc??? ND
  • Scopus 2
  • ???jsp.display-item.citation.isi??? 8
social impact