Research map / Label-efficient and robust learning / Domain adaptation and generalization
Learning to Discover Cross-Domain Relations with Generative Adversarial Networks
Read the paperCheck your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Couple both domain mappings and reconstruct inputs from both domains | DiscoGAN · Toy 2D Gaussian-mixture domains | mode coverage and mode collapse | Helped |
| Use a single GAN mapping without reconstruction loss (ablation) | Standard GAN · Toy 2D Gaussian-mixture domains | mode collapse | Hurt |
| Add reconstruction loss but reconstruct in only one direction (ablation) | GAN with reconstruction loss · Toy 2D Gaussian-mixture domains | mode collapse | Mixed |
| Add reconstruction loss but constrain only one direction (ablation) | GAN with reconstruction loss · Toy 2D Gaussian-mixture domains | mode coverage | Hurt |
| Couple bidirectional mappings with reconstruction in both directions | DiscoGAN · Car-to-car translation | correlation between input and translated azimuth | Helped |
| Use a standard GAN without bidirectional reconstruction (ablation) | Standard GAN · Car-to-car translation | azimuth preservation | Hurt |
| Add reconstruction loss but retain only a one-directional constraint (ablation) | GAN with reconstruction loss · Car-to-car translation | azimuth preservation | Hurt |
| Couple bidirectional mappings with reconstruction in both directions | DiscoGAN · Face-to-face translation | rotation variation | Helped |
| Use a standard GAN without bidirectional reconstruction (ablation) | Standard GAN · Face-to-face translation | rotation variation | Hurt |
| Add reconstruction loss but retain only a one-directional constraint (ablation) | GAN with reconstruction loss · Face-to-face translation | rotation variation | Hurt |
Closest papers by meaning
- JointGAN: Multi-Domain Joint Distribution Learning with Generative Adversarial Nets (ICML 2018 · similarity 0.86)
- Unsupervised Cross-Domain Image Generation (ICLR 2017 · similarity 0.85)
- Generate to Adapt: Aligning Domains Using Generative Adversarial Networks (CVPR 2018 · similarity 0.85)
- Domain Invariant Representation Learning with Domain Density Transformations (NeurIPS 2021 · similarity 0.84)
- Coupled Generative Adversarial Networks (NeurIPS 2016 · similarity 0.84)
- Learning Generative Models across Incomparable Spaces (ICML 2019 · similarity 0.84)
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.