Research map / 3D vision / Depth and stereo
Dense Metric Depth Completion from Sparse Direct Time-of-Flight Sensors
Hakyeong Kim, Ruicheng Wang, Chengtang Yao, Jiaolong Yang, Min H. Kim
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 directional mask to joint attention (ablation) | dual-branch ViT-S encoder · Architecture ablation (average of 6 datasets) | Rel (%): 3.87 -> 3.46 | Improved |
| add directional mask to joint attention (ablation) | dual-branch ViT-S encoder · Architecture ablation (average of 6 datasets) | delta accuracy (%): 75.6 -> 77.9 | Improved |
| fuse depth in the encoder rather than decoder-level depth prompting (ablation) | dual-branch ViT-S encoder · Architecture ablation (average of 6 datasets) | Rel (%): 4.07 -> 3.87 | Improved |
| fuse with masked joint attention instead of decoder depth prompting (ablation) | dual-branch ViT-S encoder · Architecture ablation (average of 6 datasets) | delta accuracy (%): 73.6 -> 77.9 | Improved |
| scale backbone from ViT-S to ViT-B | Ours · Architecture ablation (average of 6 datasets) | Rel (%): 3.46 -> 3.28 | Improved |
| scale backbone from ViT-S to ViT-L | Ours · Architecture ablation (average of 6 datasets) | Rel (%): 3.46 -> 3.07 | Improved |
| scale backbone from ViT-B to ViT-L | Ours · Architecture ablation (average of 6 datasets) | delta accuracy (%): 79.0 -> 80.1 | Improved |
| compare Ours with OMNI-DC on inference speed and memory | Ours vs OMNI-DC · Zero-shot depth completion (6 datasets) | inference time: 638 ms -> 34 ms | Improved |
| compare Ours with OMNI-DC on accuracy | Ours vs OMNI-DC · KITTI-DC (zero-shot) | Rel (%): 1.48 -> 2.00 | Hurt |
| compare Ours with OMNI-DC on accuracy | Ours vs OMNI-DC · ETH3D (zero-shot) | Rel (%): 1.05 -> 0.84 | Improved |
| compare Ours with OMNI-DC under spatial shifting noise | Ours vs OMNI-DC · Noise robustness: spatial shifting (average of 4 datasets) | Rel (%): 2.78 -> 2.12 | Improved |
| compare Ours with PromptDA ViT-L despite smaller backbone | Ours (ViT-S) vs PromptDA (ViT-L) · Zero-shot depth completion (average of 6 datasets) | Rel (%): 5.14 -> 3.46 | Improved |
Closest papers by meaning
- SVDC: Consistent Direct Time-of-Flight Video Depth Completion with Frequency Selective Fusion (CVPR 2025 · similarity 0.90)
- DEPTHOR: Depth Enhancement from a Practical Light-Weight dToF Sensor and RGB Image (ICCV 2025 · similarity 0.90)
- Learnable Burst-Encodable Time-of-Flight Imaging for High-Fidelity Long-Distance Depth Sensing (NeurIPS 2025 · similarity 0.87)
- Zero-shot Depth Completion via Test-time Alignment with Affine-invariant Depth Prior (AAAI 2025 · similarity 0.86)
- Flexible Depth Completion for Sparse and Varying Point Densities (CVPR 2024 · similarity 0.86)
- PatchFusion: An End-to-End Tile-Based Framework for High-Resolution Monocular Metric Depth Estimation (CVPR 2024 · 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.
