Research map / Point clouds / Point cloud classification
Unsupervised Depth Completion With Calibrated Backprojection Layers
Alex Wong, Stefano Soatto
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 the sparse-to-dense module (ablation) | KBNet · KITTI validation | MAE: 287.76 -> 260.44 | Improved |
| add calibrated backprojection layers instead of VGG blocks (ablation) | KBNet · KITTI validation | MAE: 285.97 -> 260.44 | Improved |
| use the learned S2D instead of scaffolding (ablation) | KBNet · KITTI validation | MAE: 275.56 -> 260.44 | Improved |
| use the learned S2D instead of spatial pyramid pooling (ablation) | KBNet · KITTI validation | MAE: 273.08 -> 260.44 | Improved |
| add scaffolding to VOICED (ablation) | VOICED · KITTI validation | MAE: 347.14 -> 305.06 | Improved |
| add S2D without calibrated layers vs VOICED (sparse geometry still helps) (ablation) | KBNet w/o S2D · KITTI validation | MAE: 305.06 -> 287.76 | Improved |
| decrease sparse depth density from 0.50% to 0.15% to 0.05% (ablation) | KBNet · VOID test | MAE: 39.80 -> 77.70 -> 131.54 | Hurt |
| decrease sparse density to 0.05% (vs ScaffNet) (ablation) | KBNet · VOID test (0.05% density) | RMSE: 263.54 vs 255.08 | Hurt |
| decrease input focal length by 20 to 25% (ablation) | KBNet · KITTI validation | error | Hurt |
| increase input focal length up to 20% (ablation) | KBNet · VOID test | error | No clear effect |
| train on VOID and test on NYUv2 instead of training on NYUv2 (ablation) | KBNet · NYUv2 test (sensor transfer) | MAE: 105.76 -> 117.18 | Hurt |
| train on VOID and test on NYUv2 instead of training on NYUv2 | ScaffNet · NYUv2 test (sensor transfer) | MAE: 117.49 -> 155.20 | Hurt |
Closest papers by meaning
- Deep Depth Completion from Extremely Sparse Data: A Survey (TPAMI 2022 · similarity 0.89)
- Flexible Depth Completion for Sparse and Varying Point Densities (CVPR 2024 · similarity 0.87)
- Masked Spatial Propagation Network for Sparsity-Adaptive Depth Refinement (CVPR 2024 · similarity 0.87)
- Improving Depth Completion via Depth Feature Upsampling (CVPR 2024 · similarity 0.87)
- Distilling Monocular Foundation Model for Fine-grained Depth Completion (CVPR 2025 · similarity 0.86)
- Bilateral Propagation Network for Depth Completion (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.
