Research map / Learning theory and algorithms / Optimal transport
Experts-Guided Unbalanced Optimal Transport for ISP Learning from Unpaired and/or Paired Data
Georgy Perevozchikov, Nancy Mehta, Egor Ershov, Radu Timofte
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 |
|---|---|---|---|
| add the expert committee to UOT (ablation) | EGUOT · ZRR (Restormer, unpaired) | PSNR: 19.15 -> 21.69 | Improved |
| replace hinge GAN loss with UOT loss (ablation) | EGUOT · ZRR (Restormer, unpaired) | PSNR: 19.23 -> 21.69 | Improved |
| add the color expert (ablation) | EGUOT · ZRR (Restormer, unpaired) | Delta E: 14.8 -> 9.93 | Improved |
| add the structure expert (ablation) | EGUOT · ZRR (Restormer, unpaired) | PSNR: 20.16 -> 21.69 | Improved |
| add the frequency expert (ablation) | EGUOT · ZRR (Restormer, unpaired) | PSNR: 20.95 -> 21.69 | Improved |
| use nonzero content cost tau (ablation) | EGUOT · ZRR (Restormer, unpaired) | PSNR | Improved |
| train unpaired with EGUOT instead of other unpaired methods | EGUOT · ZRR (Restormer) | PSNR: 19.07 -> 21.69 | Improved |
| train with EGUOT in paired mode instead of original loss | MW-ISP · Mobile AI | PSNR: 24.31 -> 24.79 | Improved |
| train Restormer unpaired with EGUOT instead of paired | Restormer · Mobile AI | PSNR: 23.99 -> 22.51 | Hurt |
| train with EGUOT unpaired instead of original paired | Restormer · ZRR | PSNR: 21.51 -> 21.69 | No clear effect |
| train cmKAN with unpaired EGUOT instead of paired | cmKAN · ISPIW | PSNR: 24.22 -> 24.21 | No clear effect |
| add 15% outliers to training data (GAN baseline) (ablation) | C1 GAN · ZRR | PSNR: 19.23 -> 18.65 | Hurt |
Closest papers by meaning
- Unbalancedness in Neural Monge Maps Improves Unpaired Domain Translation (ICLR 2024 · similarity 0.88)
- Residual-Conditioned Optimal Transport: Towards Structure-Preserving Unpaired and Paired Image Restoration (ICML 2024 · similarity 0.85)
- Unsupervised Learning for Optimal Transport plan prediction between unbalanced graphs (NeurIPS 2025 · similarity 0.85)
- Overcoming Spurious Solutions in Semi-Dual Neural Optimal Transport: A Smoothing Approach for Learning the Optimal Transport Plan (ICML 2025 · similarity 0.85)
- Slight Corruption in Pre-training Data Makes Better Diffusion Models (NeurIPS 2024 · similarity 0.85)
- Your contrastive learning problem is secretly a distribution alignment problem (NeurIPS 2024 · 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.
