Research map / 3D vision / Depth and stereo
Robust Incremental Structure-from-Motion with Hybrid Features
Shaohui Liu, Yidan Gao, Tianyi Zhang, Rémi Pautrat, Johannes L Schönberger, Viktor Larsson, 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 lines and vanishing points to incremental SfM (hybrid features) | COLMAP point-based SfM · Hypersim (SIFT+NN) | AUC@1deg: 71.3 -> 82.1 | Improved |
| add hybrid features on weakly textured ETH3D scenes | COLMAP point-based SfM · ETH3D (SIFT+NN) | valid registration: 46.4% -> 59.4% | Improved |
| add hybrid features on well-textured outdoor PhotoTourism scenes | COLMAP point-based SfM · PhotoTourism IMC2020 | AUC@3deg: 24.2 -> 24.6 (N=5); 80.3 -> 82.4 (N=25); AUC@1deg 11.4 -> 11.3 at N=5 | Mixed |
| run full hybrid SfM instead of COLMAP followed by LIMAP post-refinement BA | hybrid SfM · ETH3D | valid registration: 47.6% -> 59.4% | Improved |
| add hybrid point-line BA without caching to point BA (ablation) | hybrid SfM · Hypersim | AUC@1deg: 71.3 -> 75.9 | Improved |
| add inactive support caching (ablation) | hybrid SfM · Hypersim | AUC@1deg: 75.9 -> 79.2 | Improved |
| add two-step refinement (ablation) | hybrid SfM · Hypersim | AUC@1deg: 79.2 -> 80.3 | Improved |
| add retriangulation of long 2D lines (ablation) | hybrid SfM · Hypersim | AUC@1deg: 80.3 -> 80.8 | Improved |
| add VP associations (ablation) | hybrid SfM · Hypersim | AUC@1deg: 80.8 -> 81.2 | Improved |
| add hybrid registration with line and VP correspondences (ablation) | hybrid SfM · ETH3D | valid registration: 47.6% -> 59.0% | Improved |
| add uncertainty-aware reweighting in registration (ablation) | hybrid SfM · ETH3D | valid registration: 59.0% -> 59.4% | Improved |
| add uncertainty-aware reweighting to point-alone localization | localization pipeline · Cambridge | median error cm: 7.1 -> 6.4 | Improved |
Closest papers by meaning
- VGGSfM: Visual Geometry Grounded Deep Structure From Motion (CVPR 2024 · similarity 0.90)
- MP-SfM: Monocular Surface Priors for Robust Structure-from-Motion (CVPR 2025 · similarity 0.89)
- Dense-SfM: Structure from Motion with Dense Consistent Matching (CVPR 2025 · similarity 0.88)
- Detector-Free Structure from Motion (CVPR 2024 · similarity 0.88)
- Structure-from-Motion with a Non-Parametric Camera Model (CVPR 2025 · similarity 0.87)
- Leveraging Camera Triplets for Efficient and Accurate Structure-from-Motion (CVPR 2024 · similarity 0.87)
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.
