Research map / Image generation / Super-resolution
Robust Self-Supervised Cross-Modal Super-Resolution against Real-World Misaligned Observations
Xiaoyu Dong, Jiahuan Li, Ziteng Cui, Naoto Yokoya
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 misalignment-aware feature translator (ablation) | RobSelf-De · Real-world RGB-depth x2 | RMSE: 2.55 -> 2.37 | Improved |
| add content-aware reference filter (ablation) | RobSelf-De · Real-world RGB-depth x2 | RMSE: 2.55 -> 2.31 | Improved |
| combine translator and filter (ablation) | RobSelf-De · Real-world RGB-depth x2 | RMSE: 2.55 -> 2.18 | Improved |
| joint translation-based alignment vs MINIMA pre-alignment two-stage pipelines | RobSelf · Synthesized depth SR x4 | RMSE: 1.88 (MMSR+pre-align) -> 1.43 | Improved |
| joint alignment vs pre-alignment on real-world data | RobSelf · Real-world RGB-depth x2 | RMSE: 2.74 (SSGNet+pre-align) -> 2.18 | Improved |
| add pre-alignment to existing baselines | P2P · Synthesized depth SR x4 | RMSE: 2.81 -> 2.33 | Improved |
| add pre-alignment to DCTNet | DCTNet · Synthesized depth SR x8 | RMSE: 3.34 -> 3.33 | No clear effect |
| add pre-alignment on real-world RGB-NIR | P2P · Real-world NIR x2 | RMSE: 18.72 -> 18.91 | Hurt |
| pre-alignment on NIR x4 NIQE | P2P · Real-world NIR x4 | NIQE: 6.32 -> 5.32; P2P best, ahead of RobSelf 7.60 | Improved |
| RobSelf-Re vs P2P efficiency | RobSelf-Re · Task III x2 | runtime: 982s -> 64s | Improved |
| deformable conv alignment (De) vs simple resampling (Re) | RobSelf · Depth SR x4 | RMSE: 1.52 -> 1.43 | Improved |
Closest papers by meaning
- Learning Mutual Modulation for Self-Supervised Cross-Modal Super-Resolution (ECCV 2022 · similarity 0.91)
- Single Pair Cross-Modality Super Resolution (CVPR 2021 · similarity 0.89)
- Low-Res Leads the Way: Improving Generalization for Super-Resolution by Self-Supervised Learning (CVPR 2024 · similarity 0.85)
- SRSR: Enhancing Semantic Accuracy in Real-World Image Super-Resolution with Spatially Re-Focused Text-Conditioning (NeurIPS 2025 · similarity 0.85)
- Towards Realistic Data Generation for Real-World Super-Resolution (ICLR 2025 · similarity 0.85)
- SSHNet: Unsupervised Cross-modal Homography Estimation via Problem Reformulation and Split Optimization (CVPR 2025 · 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.
