Research map / Label-efficient and robust learning / Domain adaptation and generalization
NAM: Non-Adversarial Unsupervised Domain Mapping
Yedid Hoshen, Lior Wolf
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use NAM (fixed pre-trained generator, no adversarial or cycle loss) instead of CycleGAN | CycleGAN · SVHN to MNIST translation | translated digit classification accuracy (%): 26.8 -> 33.3 | Improved |
| use NAM instead of CycleGAN in the harder information-gaining direction | CycleGAN · MNIST to SVHN translation | translated digit classification accuracy (%): 17.7 -> 31.9 | Improved |
| use NAM instead of DiscoGAN | DiscoGAN · Car2Car | root median residual deviation from linear alignment: 13.81 -> 1.47 | Improved |
| use a stronger pre-trained generator (DCGAN, Spectral-Norm GAN) instead of VAE or GLO (ablation) | NAM · Edges2Shoes | qualitative translation quality | Improved |
| add VGG perceptual loss to reconstruction loss (ablation) | NAM · image domain mapping | qualitative analogy quality | Improved |
| add perceptual loss on Car2Car (versus plain pixel loss) (ablation) | NAM · Car2Car | orientation alignment | No clear effect |
| apply NAM to car2head mapping | NAM · Car2Head | orientation relation | Hurt |
| map from X to Y (T maps generated image to source) rather than Y to X | NAM · design choice | mode collapse | Improved |
Closest papers by meaning
- Domain-Scalable Unpaired Image Translation via Latent Space Anchoring (TPAMI 2023 · similarity 0.88)
- Non-Adversarial Mapping with VAEs (NeurIPS 2018 · similarity 0.88)
- Unsupervised Image-to-Image Translation With Generative Prior (CVPR 2022 · similarity 0.88)
- Unpaired Image-To-Image Translation Using Cycle-Consistent Adversarial Networks (ICCV 2017 · similarity 0.88)
- Benign Examples: Imperceptible Changes Can Enhance Image Translation Performance (AAAI 2020 · similarity 0.88)
- Unsupervised Pixel-Level Domain Adaptation With Generative Adversarial Networks (CVPR 2017 · similarity 0.87)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
