Research map / 3D vision / Depth and stereo
Adaptive Unimodal Cost Volume Filtering for Deep Stereo Matching
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 direct unimodal cost-volume supervision (ablation) | PSMNet · Scene Flow ALL | EPE [px]: 1.101 -> 0.920 | Helped |
| adapt per-pixel variance instead of using a uniform unimodal distribution (ablation) | AcfNet · Scene Flow ALL | EPE [px]: 0.920 -> 0.867 | Helped |
| adapt per-pixel variance instead of using a uniform unimodal distribution (ablation) | AcfNet · Scene Flow OCC | EPE [px]: 2.996 -> 2.736 | Helped |
| replace cross-entropy loss with stereo focal loss (ablation) | AcfNet with fixed variance σ=1.2 · Scene Flow | EPE [px]: 0.965 -> 0.920 | Helped |
| set the variance sensitivity to s=1 rather than other tested values (ablation) | AcfNet · Scene Flow | EPE | Helped |
| tune the confidence-loss weight to λ_confidence=8.0 (ablation) | AcfNet · Scene Flow | EPE | Helped |
| add adaptive unimodal cost-volume filtering (ablation) | GC-Net · Scene Flow ALL | EPE [px]: 0.871 -> 0.822 | Helped |
Closest papers by meaning
- CFNet: Cascade and Fused Cost Volume for Robust Stereo Matching (CVPR 2021 · similarity 0.90)
- CBMV: A Coalesced Bidirectional Matching Volume for Disparity Estimation (CVPR 2018 · similarity 0.90)
- Multi-Level Context Ultra-Aggregation for Stereo Matching (CVPR 2019 · similarity 0.90)
- Stereo Risk: A Continuous Modeling Approach to Stereo Matching (ICML 2024 · similarity 0.90)
- Cascade Cost Volume for High-Resolution Multi-View Stereo and Stereo Matching (CVPR 2020 · similarity 0.90)
- Modeling Stereo-Confidence out of the End-to-End Stereo-Matching Network via Disparity Plane Sweep (AAAI 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.