Research map / Image generation / Super-resolution
ResShift: Efficient Diffusion Model for Image Super-resolution by Residual Shifting
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 |
|---|---|---|---|
| shift residual from HR to LR image instead of Gaussian-noise diffusion (15 steps) | LDM-15 vs ResShift · ImageNet-Test x4 SR | LPIPS: 0.269 -> 0.231 | Helped |
| shift residual from HR to LR image instead of Gaussian-noise diffusion | LDM-15 · ImageNet-Test x4 SR | PSNR (dB): 24.89 -> 25.01 | Helped |
| use ResShift 15 steps instead of LDM-100 | LDM-100 · ImageNet-Test x4 SR | runtime (s): 0.413 -> 0.105 | Helped |
| increase diffusion steps T from 10 to 50 (ablation) | ResShift · ImageNet-Test x4 SR | CLIPIQA: 0.5492 -> 0.6489 | Helped |
| increase diffusion steps T from 10 to 50 (ablation) | ResShift · ImageNet-Test x4 SR | PSNR (dB): 25.20 -> 24.22 | Hurt |
| increase p from 0.3 to 3.0 (ablation) | ResShift T=15 · ImageNet-Test x4 SR | CLIPIQA: 0.5922 -> 0.4041 | Hurt |
| increase p from 0.3 to 3.0 (ablation) | ResShift T=15 · ImageNet-Test x4 SR | PSNR (dB): 25.01 -> 25.39 | Helped |
| increase kappa from 2.0 to 16.0 (ablation) | ResShift T=15 · ImageNet-Test x4 SR | CLIPIQA: 0.5922 -> 0.4898 | Hurt |
| decrease kappa from 2.0 to 0.5 (ablation) | ResShift T=15 · ImageNet-Test x4 SR | CLIPIQA: 0.5922 -> 0.5700 | Hurt |
| drop weight w_t from training loss (ablation) | ResShift · ImageNet-Test x4 SR | quality | Helped |
| ResShift vs BSRGAN on real-world MUSIQ | ResShift · RealSR | MUSIQ: 63.586 -> 59.873 | Hurt |
Closest papers by meaning
- Taming Diffusion Prior for Image Super-Resolution with Domain Shift SDEs (NeurIPS 2024 · similarity 0.94)
- SinSR: Diffusion-Based Image Super-Resolution in a Single Step (CVPR 2024 · similarity 0.94)
- Arbitrary-steps Image Super-resolution via Diffusion Inversion (CVPR 2025 · similarity 0.93)
- Solving Diffusion ODEs with Optimal Boundary Conditions for Better Image Super-Resolution (ICLR 2024 · similarity 0.92)
- Efficient Conditional Diffusion Model with Probability Flow Sampling for Image Super-resolution (AAAI 2024 · similarity 0.91)
- ResDiff: Combining CNN and Diffusion Model for Image Super-resolution (AAAI 2024 · similarity 0.91)
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.