Research map / 3D vision / Depth and stereo
Robust Frame-to-Frame Camera Rotation Estimation in Crowded Scenes
Fabien Delattre, David Dirnfeld, Phat Nguyen, Stephen K Scarano, Michael J Jones, Pedro Miraldo, Erik Learned-Miller
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 Hough-style voting on SO(3) instead of RANSAC-robustified baselines | camera rotation estimation · BUSS (17 crowded sequences) | rotation error (deg): 0.16 -> 0.12 (vs best RANSAC baseline H&J 500 iters) | Improved |
| add RANSAC to baselines in crowded scenes | Heeger & Jepson · BUSS (17 crowded sequences) | rotation error (deg): 0.25 -> 0.16 | Improved |
| add RANSAC to baselines in mostly static scenes | continual baselines · IRSTV (mostly static) | rotation error (deg): 0.15 -> 0.12 (Heeger & Jepson, 500 iters) | No clear effect |
| use proposed method on static scenes versus best baseline | camera rotation estimation · IRSTV (mostly static) | rotation error (deg): 0.14 vs 0.12 (Bruss & Horn) | No clear effect |
| use Longuet-Higgins motion model instead of perspective projection (ablation) | Hough rotation voting · BUSS (17 crowded sequences) | run time at small bin sizes | Improved |
| reduce bin size on SO(3) too far (ablation) | Hough rotation voting · BUSS (17 crowded sequences) | rotation error (deg) | Hurt |
| vary spatial sampling step of optical flow from 1 to 80 (ablation) | Hough rotation voting · BUSS (17 crowded sequences) | rotation error (deg): 0.11 to 0.13 degrees | No clear effect |
| use continual methods instead of discrete methods | camera rotation baselines · BUSS and IRSTV | rotation error (deg): 0.22 (P&V) vs 0.33 (Nister 50K iters) on BUSS | Improved |
Closest papers by meaning
- Dense Dynamic Scene Reconstruction and Camera Pose Estimation from Multi-View Videos (ECCV 2026 · similarity 0.89)
- MegaSaM: Accurate, Fast and Robust Structure and Motion from Casual Dynamic Videos (CVPR 2025 · similarity 0.88)
- Shape of Motion: 4D Reconstruction from a Single Video (ICCV 2025 · similarity 0.87)
- Zero-Shot Monocular Scene Flow Estimation in the Wild (CVPR 2025 · similarity 0.86)
- Cycle-Sync: Robust Global Camera Pose Estimation through Enhanced Cycle-Consistent Synchronization (NeurIPS 2025 · similarity 0.85)
- Extreme Rotation Estimation in the Wild (CVPR 2025 · similarity 0.85)
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.
