Research map / 3D vision / Depth and stereo
Rethinking Depth Estimation for Multi-View Stereo: A Unified Representation
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 regression with unification representation (ablation) | CasMVSNet · DTU evaluation set | Overall (mm): 0.340 -> 0.327 | Helped |
| Use UFL instead of BCE (ablation) | Unification on CasMVSNet · DTU evaluation set | Overall (mm): 0.327 -> 0.320 | Helped |
| Use UFL instead of GFL (ablation) | Unification on CasMVSNet · DTU evaluation set | Overall (mm): 0.325 -> 0.320 | Helped |
| Add adaptive aggregation (ablation) | Unification + UFL on CasMVSNet · DTU evaluation set | Overall (mm): 0.320 -> 0.317 | Helped |
| Add finer ground-truth (ablation) | Unification + UFL + adaptive aggregation · DTU evaluation set | Overall (mm): 0.317 -> 0.315 | Helped |
| Use regression finetuning after classification (ablation) | CasMVSNet classification baseline · DTU evaluation set | Overall (mm) | Helped |
| Use UFL with 50% of training data (ablation) | Unification on CasMVSNet · DTU evaluation set | Overall (mm) | Mixed |
| Use dedicated-function base 5 rather than e (ablation) | UFL · DTU evaluation set | Overall (mm): 0.318 -> 0.317 | Helped |
| Expand positive scaling range from [1, 2) to [1, 3) (ablation) | UFL with base 5 · DTU evaluation set | Overall (mm): 0.317 -> 0.315 | Helped |
| Use stage-wise UFL parameters rather than identical alpha-minus and lambda values (ablation) | Unification + UFL · DTU evaluation set | Overall (mm): 0.336 -> 0.320 | Helped |
| Regress proximity rather than offset directly (ablation) | Unification · DTU evaluation set | Overall (mm): 0.383 -> 0.327 | Helped |
| Use unification rather than regression (ablation) | CasMVSNet · DTU evaluation set | Overall (mm): 0.340 -> 0.327 | Helped |
Closest papers by meaning
- Efficient Edge-Preserving Multi-View Stereo Network for Depth Estimation (AAAI 2023 · similarity 0.91)
- Single View Stereo Matching (CVPR 2018 · similarity 0.90)
- Generalized Binary Search Network for Highly-Efficient Multi-View Stereo (CVPR 2022 · similarity 0.90)
- Learning Inverse Depth Regression for Multi-View Stereo with Correlation Cost Volume (AAAI 2020 · similarity 0.90)
- MVSAnywhere: Zero-Shot Multi-View Stereo (CVPR 2025 · similarity 0.88)
- A Global Depth-Range-Free Multi-View Stereo Transformer Network with Pose Embedding (NeurIPS 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.