Research map / Image generation / Super-resolution
Learned Dynamic Guidance for Depth Image Reconstruction
Shuhang Gu, Shi Zeng Guo, Wangmeng Zuo, Yunjin Chen, Radu Timofte, Luc Van Gool, Lei Zhang
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 zero-mean constraint on filters (ablation) | DG-RBF · Middlebury x8 (random init) | RMSE: 3.22 -> 3.00 | Improved |
| add zero-mean constraint on filters (model init) (ablation) | DG-RBF · Middlebury x8 (model init) | RMSE: 3.23 -> 2.25 | Improved |
| use model-inspired instead of random init for filters (ablation) | DG-RBF · Middlebury x8 | RMSE: 3.00 -> 2.25 | Improved |
| use model-inspired instead of random init for penalty functions (ablation) | DG-RBF · Middlebury x8 | RMSE: 2.37 -> 2.25 | Improved |
| increase filter size from 5x5 to 9x9 (24 filters) (ablation) | DG-RBF · Middlebury x8 | RMSE: 2.45 -> 2.18 | Improved |
| increase filter number from 12 to 48 (9x9) (ablation) | DG-RBF · Middlebury x8 | RMSE: 2.28 -> 2.14 | Improved |
| increase filter size from 9x9 to 11x11 (48 filters) (ablation) | DG-RBF · Middlebury x8 | RMSE: 2.14 -> 2.13 | No clear effect |
| use scaling factor 20 instead of 10 (33 kernels) (ablation) | DG-RBF · Middlebury x8 | RMSE: 2.18 -> 2.23 | Hurt |
| increase DG-RBF stages from 1 to 5 (ablation) | DG-RBF · Middlebury x8 | RMSE: 2.88 -> 2.18 | Improved |
| increase DG-RBF stages from 5 to 8 (ablation) | DG-RBF · Middlebury x8 | RMSE: 2.18 -> 2.19 | No clear effect |
| increase DG-CNN stages from 1 to 3 (ablation) | DG-CNN · Middlebury x8 | RMSE: 1.57 -> 1.35 | Improved |
| use stage-wise residual learning instead of global residual only (multi-stage) (ablation) | DG-CNN · Middlebury x8 | RMSE: 1.53 -> 1.35 | Improved |
Closest papers by meaning
- Learning Dynamic Guidance for Depth Image Enhancement (CVPR 2017 · similarity 0.93)
- Guided Depth Super-Resolution by Deep Anisotropic Diffusion (CVPR 2023 · similarity 0.87)
- Improving Depth Completion via Depth Feature Upsampling (CVPR 2024 · similarity 0.84)
- SGNet: Structure Guided Network via Gradient-Frequency Awareness for Depth Map Super-resolution (AAAI 2024 · similarity 0.84)
- RGB-Only Supervised Camera Parameter Optimization in Dynamic Scenes (NeurIPS 2025 · similarity 0.84)
- DORNet: A Degradation Oriented and Regularized Network for Blind Depth Super-Resolution (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.
