DualGAN: Unsupervised Dual Learning for Image-to-Image Translation

Yi, Zili; Zhang, Hao; Tan, Ping; Gong, Maoguo

doi:10.48550/arxiv.1704.02510

Cited by 105 publications

(78 citation statements)

References 1 publication

Supporting

Mentioning

Contrasting

Order By: Relevance

“…But it is not suitable for tasks where paired images are not available. To alleviate the burden of obtaining data pairs, unsupervised image-to-image translation approaches have been proposed [7], [33], [34], [35], which resort to the cycle consistency constraint for additional supervision. Similar ideas can also be found in [35], [36], [37].…”

Section: Related Workmentioning

confidence: 99%

A Domain Gap Aware Generative Adversarial Network for Multi-domain Image Translation

Xu,

Wang

2021

Preprint

View full text Add to dashboard Cite

Recent image-to-image translation models have shown great success in mapping local textures between two domains. Existing approaches rely on a cycle-consistency constraint that supervises the generators to learn an inverse mapping. However, learning the inverse mapping introduces extra trainable parameters and it is unable to learn the inverse mapping for some domains. As a result, they are ineffective in the scenarios where (i) multiple visual image domains are involved; (ii) both structure and texture transformations are required; and (iii) semantic consistency is preserved. To solve these challenges, the paper proposes a unified model to translate images across multiple domains with significant domain gaps. Unlike previous models that constrain the generators with the ubiquitous cycle-consistency constraint to achieve the content similarity, the proposed model employs a perceptual self-regularization constraint. With a single unified generator, the model can maintain consistency over the global shapes as well as the local texture information across multiple domains. Extensive qualitative and quantitative evaluations demonstrate the effectiveness and superior performance over state-of-the-art models. It is more effective in representing shape deformation in challenging mappings with significant dataset variation across multiple domains.

show abstract

Section: Related Workmentioning

confidence: 99%

A Domain Gap Aware Generative Adversarial Network for Multi-domain Image Translation

Xu,

Wang

2021

Preprint

View full text Add to dashboard Cite

show abstract

“…The existing solutions often assume a one to one mapping between the domains, i.e., that there exists a function y such that given a sample a in domain A, maps it to an analog sample in domain B. In fact, the circularity based constraints by ; ; Yi et al (2017) are based on this assumption, since going from one domain to the other and back, it is assumed that the original sample is obtained, which requires no loss of information. However, to employ an example made popular by , when going from a zebra to a horse, the stripes are lost, which results in an ambiguity when mapping in the other direction.…”

Section: Previous Workmentioning

confidence: 99%

“…In the problem of unsupervised domain translation, the algorithm receives two sets of samples, one from each domain, and learns a function that maps between a sample in one domain to the analogous sample in the other domain Yi et al, 2017;Benaim & Wolf, 2017;Liu & Tuzel, 2016;Liu et al, 2017;Choi et al, 2017;Conneau et al, 2017;Zhang et al, 2017a;b;Lample et al, 2018). The term unsupervised means, in this context, that the two sets are unpaired.…”

Section: Introductionmentioning

confidence: 99%

Emerging Disentanglement in Auto-Encoder Based Unsupervised Image Content Transfer

Press,

Galanti,

Benaim

et al. 2020

Preprint

View full text Add to dashboard Cite

We study the problem of learning to map, in an unsupervised way, between domains A and B, such that the samples b ∈ B contain all the information that exists in samples a ∈ A and some additional information. For example, ignoring occlusions, B can be people with glasses, A people without, and the glasses, would be the added information. When mapping a sample a from the first domain to the other domain, the missing information is replicated from an independent reference sample b ∈ B. Thus, in the above example, we can create, for every person without glasses a version with the glasses observed in any face image. Our solution employs a single two-pathway encoder and a single decoder for both domains. The common part of the two domains and the separate part are encoded as two vectors, and the separate part is fixed at zero for domain A. The loss terms are minimal and involve reconstruction losses for the two domains and a domain confusion term. Our analysis shows that under mild assumptions, this architecture, which is much simpler than the literature guided-translation methods, is enough to ensure disentanglement between the two domains. We present convincing results in a few visual domains, such as no-glasses to glasses, adding facial hair based on a reference image, etc.

show abstract

“…This problem is inherently ill-posed, as multiple analogous solutions may exist. In a number of different approaches [39,18,33] a circularity constraint is used to reduce this ambiguity. COGAN [24] and UNIT [23] enforce a shared latent representation between the two domains.…”

Section: Previous Workmentioning

confidence: 99%

Mask Based Unsupervised Content Transfer

Mokady¹,

Benaim²,

Wolf³

et al. 2019

Preprint

View full text Add to dashboard Cite

We consider the problem of translating, in an unsupervised manner, between two domains where one contains some additional information compared to the other. The proposed method disentangles the common and separate parts of these domains and, through the generation of a mask, focuses the attention of the underlying network to the desired augmentation alone, without wastefully reconstructing the entire target. This enables state-of-the-art quality and variety of content translation, as shown through extensive quantitative and qualitative evaluation. Furthermore, the novel mask-based formulation and regularization is accurate enough to achieve state-of-the-art performance in the realm of weakly supervised segmentation, where only class labels are given. To our knowledge, this is the first report that bridges the problems of domain disentanglement and weakly supervised segmentation. Our code is publicly available at https://github.com/rmokady/mbu-content-tansfer.Preprint. Under review.

show abstract

DualGAN: Unsupervised Dual Learning for Image-to-Image Translation

Cited by 105 publications

References 1 publication

A Domain Gap Aware Generative Adversarial Network for Multi-domain Image Translation

A Domain Gap Aware Generative Adversarial Network for Multi-domain Image Translation

Emerging Disentanglement in Auto-Encoder Based Unsupervised Image Content Transfer

Mask Based Unsupervised Content Transfer

Contact Info

Product

Resources

About