Research map / Image generation / Editing and style transfer
Extremal Domain Translation with Neural Optimal Transport
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 |
|---|---|---|---|
| Increase the IT weight from w=1 to w=8 (ablation) | IT · Image translation benchmarks | ℓ² transport cost: 0.297 -> 0.094 (CelebA → anime); 0.368 -> 0.252 (Handbag → shoes); 0.603 -> 0.408 (Textures → chairs); 0.224 -> 0.196 (F | Helped |
| Increase the IT weight from w=1 to w=8 (ablation) | IT · Image translation benchmarks | FID: 14.65 -> 22.84 (CelebA → anime); 27.10 -> 53.80 (Handbag → shoes); 20.95 -> 23.67 (FFHQ → comics) | Mixed |
| Add a small identity loss (λ=50) (ablation) | CycleGAN · CelebA → anime | FID: 22.9 -> 20.8 | Helped |
| Increase CycleGAN identity-loss weight from λ=0 to λ=500 (ablation) | CycleGAN · CelebA → anime | ℓ² transport cost: 0.33 -> 0.01 | Helped |
| Increase CycleGAN identity-loss weight from λ=0 to λ=500 (ablation) | CycleGAN · CelebA → anime | FID: 22.9 -> 122.8 | Hurt |
| Add ℓ² identity loss to StarGAN-v2 (λ=10) (ablation) | StarGAN-v2 · Handbag → shoes | ℓ² transport cost: 0.739 -> 0.244 | Mixed |
| Increase StarGAN-v2 similarity-loss weight from λ=0 to λ=10 (ablation) | StarGAN-v2 · Handbag → shoes | FID: 25.45 -> 22.36 | Helped |
| Use an ℓ² similarity term in CycleGAN instead of the paper's ℓ¹ identity loss (ablation) | CycleGAN · Image translation | ℓ² transport cost | Hurt |
| Use a soft unbalanced-OT constraint instead of hard partial-OT/IT constraints (ablation) | Unbalanced OT · Wi-Fi toy example | target-mass alignment | Hurt |
| Use GAN+ℓ² or unbalanced OT for the Accept toy task (ablation) | GAN+ℓ² and unbalanced OT · Accept toy example | target-domain translation | Hurt |
Closest papers by meaning
- Overcoming Spurious Solutions in Semi-Dual Neural Optimal Transport: A Smoothing Approach for Learning the Optimal Transport Plan (ICML 2025 · similarity 0.88)
- Unbalancedness in Neural Monge Maps Improves Unpaired Domain Translation (ICLR 2024 · similarity 0.87)
- Unpaired Image-to-Image Translation With Shortest Path Regularization (CVPR 2023 · similarity 0.87)
- Unpaired Image-to-Image Translation via Neural Schrödinger Bridge (ICLR 2024 · similarity 0.85)
- Improving Neural Optimal Transport via Displacement Interpolation (ICLR 2025 · similarity 0.85)
- Universal Neural Optimal Transport (ICML 2025 · similarity 0.85)
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.