Research map / 3D vision / Depth and stereo
ElasticMVS: Learning elastic part representation for self-supervised multi-view stereopsis
Jinzhi Zhang, Ruofan Tang, Zheng Cao, Jing Xiao, Ruqi Huang, LU FANG
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 elastic part representation to propagation (replace fixed sampled locations) (ablation) | ElasticMVS · DTU | overall distance (mm): 0.455 -> 0.349 | Improved |
| add part-aware correspondence to evaluation (ablation) | ElasticMVS · DTU | overall distance (mm): 0.430 -> 0.349 | Improved |
| add part smoothness loss (replace L1-median loss) (ablation) | ElasticMVS · DTU | overall distance (mm): 0.411 -> 0.349 | Improved |
| add elastic part representation to propagation (ablation) | ElasticMVS · T&T Intermediate | F-score: 40.05 -> 57.88 | Improved |
| add part-aware correspondence to evaluation (ablation) | ElasticMVS · T&T Intermediate | F-score: 43.82 -> 57.88 | Improved |
| add part smoothness loss (ablation) | ElasticMVS · T&T Intermediate | F-score: 39.58 -> 57.88 | Improved |
| add confidence map to representation training (ablation) | elastic part representation · DTU | depth error (mm): 3.55 -> 2.52 | Improved |
| add spatial concentration loss to representation training (ablation) | elastic part representation · DTU | depth error (mm): 3.01 -> 2.52 | Improved |
| add confidence map to representation training (ablation) | elastic part representation · DTU | contrastive loss: 2.52 -> 2.28 | Improved |
| lower point-to-plane threshold epsilon from 0.02 to 0.005 (ablation) | ElasticMVS inference · DTU | accuracy / completeness (mm): 0.380 -> 0.368 acc, 0.301 -> 0.354 comp | Mixed |
| raise confidence threshold from 0.5 to 0.8 (ablation) | ElasticMVS inference · DTU | overall distance (mm): 0.441 -> 0.344 | Improved |
| raise confidence threshold from 0.7 to 0.8 (ablation) | ElasticMVS inference · DTU | overall distance (mm): 0.349 -> 0.344 | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- PatchmatchNet — PatchmatchNet: Learned Multi-View Patchmatch Stereo (CVPR 2021) · better on 4 of 5 shared results
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
- Learning a Multi-View Stereo Machine (NeurIPS 2017 · similarity 0.86)
- Learning Partonomic 3D Reconstruction from Image Collections (CVPR 2025 · similarity 0.86)
- PartSAM: A Scalable Promptable Part Segmentation Model Trained on Native 3D Data (ICLR 2026 · similarity 0.86)
- DeepMVS: Learning Multi-View Stereopsis (CVPR 2018 · similarity 0.86)
- SD-MVS: Segmentation-Driven Deformation Multi-View Stereo with Spherical Refinement and EM Optimization (AAAI 2024 · similarity 0.86)
- Large Spatial Model: End-to-end Unposed Images to Semantic 3D (NeurIPS 2024 · similarity 0.84)
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.
