NAM: Non-Adversarial Unsupervised Domain Mapping
This addresses the challenge of domain adaptation in computer vision, offering a more stable alternative to adversarial methods, though it is incremental as it builds on existing generative modeling techniques.
The paper tackled the problem of unsupervised image translation between domains without correspondences by introducing Non-Adversarial Mapping (NAM), which separates target domain generative modeling from cross-domain mapping, resulting in higher quality and resolution translations with simpler and more stable training.
Several methods were recently proposed for the task of translating images between domains without prior knowledge in the form of correspondences. The existing methods apply adversarial learning to ensure that the distribution of the mapped source domain is indistinguishable from the target domain, which suffers from known stability issues. In addition, most methods rely heavily on `cycle' relationships between the domains, which enforce a one-to-one mapping. In this work, we introduce an alternative method: Non-Adversarial Mapping (NAM), which separates the task of target domain generative modeling from the cross-domain mapping task. NAM relies on a pre-trained generative model of the target domain, and aligns each source image with an image synthesized from the target domain, while jointly optimizing the domain mapping function. It has several key advantages: higher quality and resolution image translations, simpler and more stable training and reusable target models. Extensive experiments are presented validating the advantages of our method.