Research map / Graph learning / Graph self-supervised learning
Learning Graph Regularisation for Guided Super-Resolution
Riccardo de Lutio, Alexander Becker, Stefano D'Aronco, Stefania Russo, Jan D. Wegner, Konrad Schindler
Check 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 |
|---|---|---|---|
| use learned graph regularisation with an explicit source-fidelity optimisation layer | guided depth super-resolution · Middlebury, NYUv2, and DIML at ×4, ×8, and ×16 | MSE: Middlebury: 3.60 -> 3.04 (×4), 9.48 -> 7.26 (×8), 30.6 -> 24.7 (×16); NYUv2: 6.85 -> 6.45 (×4), 24.1 -> 19.6 (×8), 84.5 | Improved |
| train on NYUv2 and retain the explicit source-fidelity constraint | Ours · DIML and Middlebury cross-dataset ×8 | low-resolution MSE: DIML: 2.4 · 10⁻³; Middlebury: 1.1 · 10⁻³ | Improved |
| replace learned features with raw Colour features (ablation) | graph regularisation for guided depth super-resolution · NYUv2 ×8 | MSE: 19.6 -> 68.4 | Hurt |
| use UEfficientNet-B2 as the graph feature extractor (ablation) | Ours · NYUv2 ×8 | MSE: 19.6 -> 24.9 | Hurt |
| use UResNet-18 instead of UResNet-50 as the graph feature extractor (ablation) | Ours · NYUv2 ×8 | MSE: 19.6 -> 21.7 | Hurt |
| use FDSR features instead of UResNet-50 features (ablation) | Ours · NYUv2 ×8 | MSE: 19.6 -> 30.4 | Hurt |
| use GDPT instead of UResNet-50 as the graph feature extractor (ablation) | Ours · NYUv2 ×8 | MSE: 19.6 -> 22.3 | Hurt |
| increase feature extractor capacity from UResNet-18 to UResNet-50 (ablation) | Ours · NYUv2 ×8 | MAE: 1.52 -> 1.42 | Improved |
Closest papers by meaning
- Guided Super-Resolution As Pixel-to-Pixel Transformation (ICCV 2019 · similarity 0.90)
- Dual Graph Regularized Deep Unfolding Network for Guided Depth Map Super-resolution (CVPR 2026 · similarity 0.89)
- Guided Depth Super-Resolution by Deep Anisotropic Diffusion (CVPR 2023 · similarity 0.88)
- Image Processing GNN: Breaking Rigidity in Super-Resolution (CVPR 2024 · similarity 0.87)
- Pixel-level and Semantic-level Adjustable Super-resolution: A Dual-LoRA Approach (CVPR 2025 · similarity 0.86)
- Unleashing the Power of One-Step Diffusion based Image Super-Resolution via a Large-Scale Diffusion Discriminator (NeurIPS 2025 · similarity 0.86)
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.
