Research map / 3D vision / Depth and stereo
DesNet: Decomposed Scale-Consistent Network for Unsupervised Depth Completion
Read the paperCheck 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 |
|---|---|---|---|
| Add DSCL to direct absolute-depth prediction (ablation) | DesNet-i · KITTI validation split | RMSE (mm): 1176.4 -> 1084.5 | Helped |
| Add GDG to direct absolute-depth prediction (ablation) | DesNet-i · KITTI validation split | RMSE (mm): 1176.4 -> 1061.8 | Helped |
| Combine DSCL and GDG (ablation) | DesNet-i · KITTI validation split | RMSE (mm): 1176.4 -> 969.3 | Helped |
| Increase the scale-factor element count from 1 to 16 (ablation) | DSCL · KITTI validation split | RMSE (mm): 1084.5 -> 1068.2 | Helped |
| Use HW scale-factor elements rather than 1 (ablation) | DSCL · KITTI validation split | RMSE (mm): 1084.5 -> 1189.4 | Hurt |
| Generate dense reference with bilinear interpolation (ablation) | GDG · KITTI validation split | RMSE (mm): 1176.4 -> 1162.7 | Helped |
| Generate dense reference with nearest interpolation (ablation) | GDG · KITTI validation split | RMSE (mm): 1176.4 -> 1148.4 | Helped |
| Use square rather than cross dilation kernel of size 3 (ablation) | GDG · KITTI validation split | RMSE (mm): 1106.6 -> 1088.3 | Helped |
| Use square dilation size 5 rather than size 3 (ablation) | GDG · KITTI validation split | RMSE (mm): 1088.3 -> 1061.8 | Helped |
| Increase square dilation size from 5 to 7 (ablation) | GDG · KITTI validation split | RMSE (mm): 1061.8 -> 1112.9 | Hurt |
| Use dense-to-sparse attention instead of conventional attention (ablation) | GDG · KITTI validation split | RMSE (mm): 1040.2 -> 1024.5 | Helped |
| Use fast dense-to-sparse attention instead of DSA (ablation) | GDG · KITTI validation split | RMSE (mm): 1024.5 -> 1061.8 | Hurt |
Closest papers by meaning
- Distilling Monocular Foundation Model for Fine-grained Depth Completion (CVPR 2025 · similarity 0.89)
- Flexible Depth Completion for Sparse and Varying Point Densities (CVPR 2024 · similarity 0.88)
- Masked Spatial Propagation Network for Sparsity-Adaptive Depth Refinement (CVPR 2024 · similarity 0.88)
- A Simple yet Universal Framework for Depth Completion (NeurIPS 2024 · similarity 0.88)
- Improving Depth Completion via Depth Feature Upsampling (CVPR 2024 · similarity 0.87)
- Mind The Edge: Refining Depth Edges in Sparsely-Supervised Monocular Depth Estimation (CVPR 2024 · similarity 0.87)
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.