Research map / 3D vision / Depth and stereo
Cascade Cost Volume for High-Resolution Multi-View Stereo and Stereo Matching
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 |
|---|---|---|---|
| Replace the single MVSNet cost volume with a cascade cost volume | MVSNet · DTU evaluation | Overall (mm): 0.551 -> 0.355 | Helped |
| Replace the single MVSNet cost volume with a cascade cost volume | MVSNet · DTU evaluation | GPU Mem (MB): 10823 -> 5345 | Helped |
| Replace the single MVSNet cost volume with a cascade cost volume | MVSNet · DTU evaluation | Run-time (s): 1.210 -> 0.492 | Helped |
| Replace the single cost volume with a cascade cost volume | PSMNet · Scene Flow | EPE: 0.887 vs. 0.721 | Helped |
| Replace the single cost volume with a cascade cost volume | GwcNet · Scene Flow | EPE: 0.765 vs. 0.649 | Helped |
| Replace the single cost volume with a cascade cost volume | GANet11 · Scene Flow | EPE: 0.950 vs. 0.900 | Helped |
| Use higher-resolution cost volumes from feature-pyramid maps instead of bilinearly upsampled features (ablation) | MVSNet-Cas3-Ups · DTU validation | Overall (mm): 0.379 to 0.355 | Helped |
| Increase cost-volume spatial resolution across cascade stages (ablation) | MVSNet-Cas3 · DTU validation | Overall (mm): 0.453 vs. 0.355 | Helped |
| Share regularization parameters across cascade stages (ablation) | MVSNet-Cas3 · DTU validation | Overall (mm): 0.4512 vs. 0.4310 | Hurt |
| Use two rather than one cascade stage (ablation) | MVSNet · DTU validation | Overall (mm): 0.5510 -> 0.4314 | Helped |
| Increase cascade stages from two to three (ablation) | MVSNet · DTU validation | Overall (mm): 0.4314 -> 0.4310 | No clear effect |
| Increase cascade stages from three to four (ablation) | MVSNet · DTU validation | Overall (mm): 0.4310 -> 0.4364 | Hurt |
Closest papers by meaning
- Efficient Multi-View Stereo by Iterative Dynamic Cost Volume (CVPR 2022 · similarity 0.91)
- CFNet: Cascade and Fused Cost Volume for Robust Stereo Matching (CVPR 2021 · similarity 0.91)
- Cost Volume Pyramid Based Depth Inference for Multi-View Stereo (CVPR 2020 · similarity 0.91)
- Learning Inverse Depth Regression for Multi-View Stereo with Correlation Cost Volume (AAAI 2020 · similarity 0.91)
- Generalized Binary Search Network for Highly-Efficient Multi-View Stereo (CVPR 2022 · similarity 0.90)
- MVSFormer++: Revealing the Devil in Transformer's Details for Multi-View Stereo (ICLR 2024 · similarity 0.88)
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.