Research map / 3D vision / Depth and stereo
MGSfM: Multi-Camera Geometry Driven Global Structure-from-Motion
Peilin Tao, Hainan Cui, Diantao Tu, Shuhan Shen
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 |
|---|---|---|---|
| use non-bilinear instead of bilinear angle-based objective (ablation) | hybrid translation averaging · KITTI-360 seq 0010 | mean position error (m): 9e1 -> 3.4 | Improved |
| use non-bilinear instead of bilinear objective (ablation) | relative translations only · KITTI-360 seq 0010 | mean position error (m): 9e1 -> 4.7 | Improved |
| use non-bilinear instead of bilinear objective (ablation) | feature tracks only with random init · KITTI-360 seq 0010 | mean position error (m): 3e2 -> 6e2 | Hurt |
| add feature tracks to relative translations (hybrid, non-bilinear) (ablation) | translation averaging · KITTI-360 seq 0010 | mean position error (m): 4.7 -> 3.4 | Improved |
| use hybrid distance-then-angle translation averaging instead of GLOMAP-style global positioning (ablation) | DMRA+MGP · KITTI-360 seq 0010 | mean position error (m): 3e2 -> 3.4 | Improved |
| use hybrid translation averaging instead of DMRA+MGP (ablation) | DMRA+MGP · KITTI Odometry seq 01 | mean position error (m): 31.7 -> 1.1 | Improved |
| add multi-camera constraints (MCSfM vs COLMAP) | COLMAP · KITTI Odometry seq 08 | mean position error (m): 11.5 -> 2.6 | Improved |
| use global multi-camera pipeline MGSfM | MCSfM · KITTI-360 runtime | speed: more than 10x faster | Improved |
| use MGSfM | GLOMAP · ETH3D-SLAM ceiling_1 | AUC@0.5m: 66.2 -> 87.4 | Improved |
| use MGSfM | MCSfM · CAMPUS (29k images) | reconstruction time (min): 401 -> 66 | Improved |
Closest papers by meaning
- MMA: Multi-Camera Based Global Motion Averaging (AAAI 2022 · similarity 0.94)
- HSfM: Hybrid Structure-from-Motion (CVPR 2017 · similarity 0.91)
- Hybrid Rotation Averaging: A Fast and Robust Rotation Averaging Approach (CVPR 2021 · similarity 0.90)
- Very Large-Scale Global SfM by Distributed Motion Averaging (CVPR 2018 · similarity 0.90)
- Learning to Filter Outlier Edges in Global SfM (CVPR 2025 · similarity 0.86)
- VGGSfM: Visual Geometry Grounded Deep Structure From Motion (CVPR 2024 · 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.
