Research map / Image generation / Image restoration
Multi-Curve Translator for High-Resolution Photorealistic Image Translation
Yuda Song, Hui Qian, Xin Du
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 |
|---|---|---|---|
| replace conv output layer with multi-curve slicing on downsampled input | GCANet · SOTS dehazing | PSNR: 25.09 -> 25.71 | Improved |
| replace conv output layer with multi-curve slicing | MSBDN · SOTS dehazing | PSNR: 28.56 -> 28.70 | No clear effect |
| replace conv output layer with multi-curve slicing | MSBDN · HazeRD dehazing (4K) | SSIM: 0.805 -> 0.840 | Improved |
| replace conv output layer with multi-curve slicing | GCANet · HazeRD dehazing (4K) | SSIM: 0.825 -> 0.810 | Hurt |
| replace conv output layer with multi-curve slicing | DPE · MIT-Adobe-5K retouching (unpaired) | PSNR: 20.92 -> 23.09 | Improved |
| replace conv output layer with multi-curve slicing | DPED · MIT-Adobe-5K retouching (paired) | SSIM: 0.886 -> 0.936 | Improved |
| replace conv output layer with multi-curve slicing | MSBDN · Dehazing 4K speed | FPS: 2.2 -> 35.9 | Improved |
| replace conv output layer with multi-curve slicing | CycleGAN · 4K speed (A100) | FPS: 2.7 -> 116.0 | Improved |
| replace conv output layer with multi-curve slicing | CycleGAN · day2dusk | FID: 89.00 -> 81.67 | Improved |
| replace conv output layer with multi-curve slicing | CycleGAN · summer2autumn | FID: 101.98 -> 103.45 | Hurt |
| replace conv output layer with multi-curve slicing | AdaIN · style transfer user study | preference: 84.2% -> 15.8% | Hurt |
| add base-output constraint during training (ablation) | MCT-CycleGAN · day2dusk | visual quality | Improved |
Closest papers by meaning
- High-Resolution Photorealistic Image Translation in Real-Time: A Laplacian Pyramid Translation Network (CVPR 2021 · similarity 0.88)
- Spatially-Adaptive Pixelwise Networks for Fast Image Translation (CVPR 2021 · similarity 0.86)
- MERIT: Multi-domain Efficient RAW Image Translation (CVPR 2026 · similarity 0.84)
- Discriminative Region Proposal Adversarial Networks for High-Quality Image-to-Image Translation (ECCV 2018 · similarity 0.84)
- Multimodal Unsupervised Image-to-image Translation (ECCV 2018 · similarity 0.84)
- High-Resolution Daytime Translation Without Domain Labels (CVPR 2020 · 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.
