Research map / 3D vision / Depth and stereo
Lightweight and Accurate Multi-View Stereo With Confidence-Aware Diffusion Model
Fangjinhua Wang, Qingshan Xu, Yew-Soon Ong, Marc Pollefeys
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 diffusion model refinement (ablation) | DiffMVS · DTU test | Overall (mm): 0.318 -> 0.308 | Improved |
| add diffusion model refinement (ablation) | DiffMVS · ETH3D training | F1-score: 70.69 -> 74.86 | Improved |
| add noise only in training (ablation) | DiffMVS · DTU test | Overall (mm): 0.318 -> 0.329 | Hurt |
| add cost volume as diffusion condition (ablation) | DiffMVS · DTU test | Overall (mm): 2.682 -> 0.308 | Improved |
| add depth context as diffusion condition (ablation) | DiffMVS · ETH3D training | F1-score: 43.07 -> 74.86 | Improved |
| add image context as diffusion condition (ablation) | DiffMVS · DTU test | Overall (mm): 0.311 -> 0.308 | Improved |
| use multiple samples with confidence-based sampling instead of a single sample (ablation) | DiffMVS · ETH3D training | F1-score: 66.77 -> 74.86 | Improved |
| add confidence-based sampling instead of no confidence (ablation) | DiffMVS · DTU test | Overall (mm): 0.319 -> 0.308 | Improved |
| use confidence for sampling rather than only regularization (ablation) | DiffMVS · ETH3D training | F1-score: 73.22 -> 74.86 | Improved |
| use GRU instead of stacked U-Nets (ablation) | DiffMVS · ETH3D training | F1-score: 71.93 -> 74.86 | Improved |
| use GRU instead of single U-Net (ablation) | DiffMVS · DTU test | Overall (mm): 2.317 -> 0.308 | Improved |
| increase DDIM sampling steps from 1 to 2 (ablation) | DiffMVS · DTU test | Depth error (mm): 4.68 -> 4.63 | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 3 on most shared results.
- IterMVS — IterMVS: Iterative Probability Estimation for Efficient Multi-View Stereo (CVPR 2022) · better on 3 of 3 shared results
- GeoMVSNet — GeoMVSNet: Learning Multi-View Stereo With Geometry Perception (CVPR 2023) · better on 0 of 1 shared result
- TransMVSNet — TransMVSNet: Global Context-Aware Multi-View Stereo Network With Transformers (CVPR 2022) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- MVDD: Multi-View Depth Diffusion Models (ECCV 2024 · similarity 0.91)
- Fast-MVSNet: Sparse-to-Dense Multi-View Stereo With Learned Propagation and Gauss-Newton Refinement (CVPR 2020 · similarity 0.91)
- Learning-Based Multi-View Stereo: A Survey (TPAMI 2026 · similarity 0.91)
- Zero-Shot Novel View and Depth Synthesis with Multi-View Geometric Diffusion (CVPR 2025 · similarity 0.90)
- Multi-view Reconstruction via SfM-guided Monocular Depth Estimation (CVPR 2025 · similarity 0.90)
- Adaptive 3D Reconstruction via Diffusion Priors and Forward Curvature-Matching Likelihood Updates (NeurIPS 2025 · similarity 0.89)
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.
