Research map / Image generation / Super-resolution
Discrete Cosine Transform Network for Guided Depth Map Super-Resolution
Check 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 |
|---|---|---|---|
| Use semi-coupled filters rather than independent filters (ablation) | DCTNet SCFE · NYU v2 test, ×4/×8/×16 | RMSE: 1.74 -> 1.59; 3.34 -> 3.16; 6.06 -> 5.84 | Helped |
| Use semi-coupled filters rather than fully-shared filters (ablation) | DCTNet SCFE · NYU v2 test, ×4/×8/×16 | RMSE: 1.80 -> 1.59; 3.48 -> 3.16; 6.46 -> 5.84 | Helped |
| Replace the DCT module with a three-layer CNN (ablation) | DCTNet · NYU v2 test, ×4/×8/×16 | RMSE: 1.59 -> 1.78; 3.16 -> 3.46; 5.84 -> 6.61 | Hurt |
| Fix λ to e^{0.1} instead of learning channel-wise parameters (ablation) | DCTNet DCT module · NYU v2 test, ×4/×8/×16 | RMSE: 1.59 -> 1.77; 3.16 -> 3.55; 5.84 -> 6.63 | Hurt |
| Remove residual connections from SCFE (ablation) | DCTNet SCFE · NYU v2 test, ×4/×8/×16 | RMSE: 1.59 -> 1.91; 3.16 -> 3.84; 5.84 -> 7.06 | Hurt |
| Increase network depth beyond P=4 or width beyond C=64 (ablation) | DCTNet · NYU v2 validation, ×4/×8/×16 | RMSE | No clear effect |
| Use DCTNet rather than the strongest listed baseline | Guided depth super-resolution · Middlebury, NYU v2, Lu, and RGBDD at ×4/×8/×16 | RMSE: Middlebury: 1.08 vs 1.10; 2.08 -> 2.05; 4.24 -> 4.19; NYU v2: 1.62 -> 1.59; 3.18 -> 3.16; 5.86 -> 5.84; Lu: 0.82 vs 0.88 | Mixed |
Closest papers by meaning
- SGNet: Structure Guided Network via Gradient-Frequency Awareness for Depth Map Super-resolution (AAAI 2024 · similarity 0.87)
- Decoupling Fine Detail and Global Geometry for Compressed Depth Map Super-Resolution (CVPR 2025 · similarity 0.86)
- DegBins: Degradation-Driven Binning for Depth Super-Resolution (NeurIPS 2026 · similarity 0.86)
- DCDepth: Progressive Monocular Depth Estimation in Discrete Cosine Domain (NeurIPS 2024 · similarity 0.86)
- Intrinsic Phase-Preserving Networks for Depth Super Resolution (AAAI 2024 · similarity 0.86)
- C2PD: Continuity-Constrained Pixelwise Deformation for Guided Depth Super-Resolution (AAAI 2025 · similarity 0.85)
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.