Research map / 3D vision / Depth and stereo
IterMVS: Iterative Probability Estimation for Efficient Multi-View Stereo
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 |
|---|---|---|---|
| combine classification with local regression for depth prediction (ablation) | IterMVS · DTU and ETH3D Training | DTU Overall (mm); ETH3D F1-score: 0.428 -> 0.363; 63.34 -> 66.36 | Helped |
| add learned confidence prediction instead of confidence from summed probabilities (ablation) | IterMVS · DTU and ETH3D Training | DTU Overall (mm); ETH3D F1-score: 0.364 -> 0.363; 66.04 -> 66.36 | Helped |
| use multi-scale features instead of single-scale features (ablation) | IterMVS · DTU and ETH3D Training | DTU Overall (mm); ETH3D F1-score: 0.370 -> 0.363; 64.02 -> 66.36 | Helped |
| use learned depth upsampling instead of bilinear upsampling (ablation) | IterMVS · DTU and ETH3D Training | DTU Overall (mm); ETH3D F1-score: 0.387 -> 0.363; 63.79 -> 66.36 | Helped |
| compute losses in inverse depth rather than depth (ablation) | IterMVS · DTU and ETH3D Training | DTU Overall (mm); ETH3D F1-score: 0.364 -> 0.363; 64.51 -> 66.36 | Mixed |
| use pixel-wise view weights instead of averaging views (ablation) | IterMVS · DTU and ETH3D Training | DTU Overall (mm); ETH3D F1-score: 0.369 -> 0.363; 60.02 -> 66.36 | Mixed |
| increase GRU iterations from 1 to 4 (ablation) | IterMVS · DTU evaluation | Overall (mm): 0.396 -> 0.363 | Helped |
| increase GRU iterations from 4 to 16 (ablation) | IterMVS · DTU evaluation | Overall (mm): 0.363 -> 0.348 | Helped |
| increase input views from 2 to 6 (ablation) | IterMVS · DTU evaluation | Overall (mm): 0.466 -> 0.362 | Helped |
| increase input views from 6 to 7 (ablation) | IterMVS · DTU evaluation | Overall (mm): 0.362 -> 0.364 | Hurt |
| train on BlendedMVS rather than DTU only | IterMVS · ETH3D Test | F1-score: 74.29 -> 80.06 | Helped |
| use IterMVS instead of PatchmatchNet | MVS on ETH3D Test · ETH3D Test | F1-score: 73.12 -> 74.29 | Helped |
Closest papers by meaning
- Efficient Multi-View Stereo by Iterative Dynamic Cost Volume (CVPR 2022 · similarity 0.89)
- GeoMVSNet: Learning Multi-View Stereo With Geometry Perception (CVPR 2023 · similarity 0.89)
- Single View Stereo Matching (CVPR 2018 · similarity 0.88)
- Any-Stereo: Arbitrary Scale Disparity Estimation for Iterative Stereo Matching (AAAI 2024 · similarity 0.88)
- MVSAnywhere: Zero-Shot Multi-View Stereo (CVPR 2025 · similarity 0.87)
- MVSFormer++: Revealing the Devil in Transformer's Details for Multi-View Stereo (ICLR 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.