Research map / 3D vision / Depth and stereo
MVSNet: Depth Inference for Unstructured Multi-view Stereo
Yao Yao, Zixin Luo, Shiwei Li, Tian Fang, Long Quan
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 depth map refinement network (ablation) | MVSNet · DTU evaluation set | f-score (<1mm): 75.58 -> 75.69 | Improved |
| add depth map refinement network (ablation) | MVSNet · DTU evaluation set | f-score (<2mm): 79.98 -> 80.25 | Improved |
| add depth map refinement network (ablation) | MVSNet · DTU validation | validation loss | No clear effect |
| use learned 2D feature extraction network instead of single 7x7 stride-4 conv layer (ablation) | MVSNet · DTU validation | validation loss | Improved |
| use variance-based cost metric instead of mean (ablation) | MVSNet · DTU validation | validation loss | Improved |
| increase test input views from 3 to 5 (trained with 3) (ablation) | MVSNet · DTU validation | validation loss | Improved |
| add more input views (N=2 vs 3) (ablation) | MVSNet · DTU validation | validation loss | Improved |
| use MVSNet instead of Gipuma on completeness | MVSNet · DTU evaluation set | completeness distance (mm): 0.873 -> 0.527 | Improved |
| use MVSNet instead of Gipuma on accuracy | MVSNet · DTU evaluation set | accuracy distance (mm): 0.283 -> 0.396 | Hurt |
| use MVSNet instead of Gipuma overall | MVSNet · DTU evaluation set | overall distance (mm): 0.578 -> 0.462 | Improved |
| apply DTU-trained MVSNet to outdoor scenes without fine-tuning | MVSNet · Tanks and Temples intermediate | mean f-score: 42.14 -> 43.48 | Improved |
Closest papers by meaning
- Fast-MVSNet: Sparse-to-Dense Multi-View Stereo With Learned Propagation and Gauss-Newton Refinement (CVPR 2020 · similarity 0.92)
- Recurrent MVSNet for High-Resolution Multi-View Stereo Depth Inference (CVPR 2019 · similarity 0.92)
- MVS2D: Efficient Multi-View Stereo via Attention-Driven 2D Convolutions (CVPR 2022 · similarity 0.91)
- Point-Based Multi-View Stereo Network (ICCV 2019 · similarity 0.91)
- Cost Volume Pyramid Based Depth Inference for Multi-View Stereo (CVPR 2020 · similarity 0.91)
- MVSCRF: Learning Multi-View Stereo With Conditional Random Fields (ICCV 2019 · similarity 0.91)
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.
