Research map / 3D vision / Depth and stereo
Robust Isometric Non-Rigid Structure-From-Motion
Shaifali Parashar, Daniel Pizarro, Adrien Bartoli
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 robust pipeline (MAD+MV+OR) on top of local normal estimators | FS-NRSfM / FR-NRSfM · Cylinder synthetic | breakdown correspondence error rate: break at about 20% -> beyond 50% | Improved |
| use FR-NRSfM (resultants) instead of FS-NRSfM (substitution) as base | robust pipeline · Cylinder synthetic | shape and depth error: RR-NRSfM performs slightly better than RS-NRSfM | Improved |
| increase image subset size from 5 to 7 with missing data (ablation) | robust pipeline · Cylinder synthetic with missing data up to 50% | reconstruction accuracy: 2.5s -> 3s | Improved |
| increase image subset size from 7 to 10 (ablation) | robust pipeline · Cylinder synthetic | accuracy vs time: 3s -> 5s | No clear effect |
| add OR step without MAD (ablation) | FS/FR-NRSfM on error-free real data · Paper, manual-Tshirt, manual-Hulk | 3D points removed: almost 1% removed without MAD; 0.1% with MAD | Mixed |
| add MAD before OR (ablation) | OR on ASIFT datasets · Rug, ASIFT-Tshirt, ASIFT-Hulk | 3D points removed: almost 15% -> fewer than 0.1% | Improved |
| add multi-view reconstruction (MVS/MVR) (ablation) | FS-NRSfM / FR-NRSfM · ASIFT-Tshirt and ASIFT-Hulk | reconstruction error | Improved |
| add MAD and MV to base estimator on error-free data (ablation) | FS-NRSfM / FR-NRSfM · Paper, manual-Tshirt, manual-Hulk | reconstruction error | No clear effect |
| add OR step (ablation) | FS-NRSfM / FR-NRSfM · error-free real data | reconstruction error | Improved |
| replace MAD with a prior outlier-rejection method (ablation) | robust pipeline · ASIFT-Tshirt / ASIFT-Hulk | mean depth error (mm): 8.3 mm and 7.5 mm to 6.7 mm and 6.2 mm | Improved |
| use MAD instead of prior outlier rejection | robust pipeline · mismatch detection time | time per image: around 500 ms -> 150-200 ms | Improved |
| use FS/FR-NRSfM instead of Pa17 | normal estimation · local normal solve | time: 1.5 s -> 7-10 ms | Improved |
Closest papers by meaning
- Isometric Non-Rigid Shape-from-Motion with Riemannian Geometry Solved in Linear Time (TPAMI 2017 · similarity 0.92)
- Self-Calibrating Isometric Non-Rigid Structure-from-Motion (ECCV 2018 · similarity 0.92)
- Isometric Non-Rigid Shape-From-Motion in Linear Time (CVPR 2016 · similarity 0.91)
- Non-Rigid Structure-from-Motion: Temporally-Smooth Procrustean Alignment and Spatially-Variant Deformation Modeling (CVPR 2024 · similarity 0.89)
- Deep Non-Rigid Structure-from-Motion Revisited: Canonicalization and Sequence Modeling (AAAI 2025 · similarity 0.88)
- Structure-from-Motion with a Non-Parametric Camera Model (CVPR 2025 · 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.
