Research map / Graph learning / Graph algorithms and clustering
NeuRoRA: Neural Robust Rotation Averaging
Pulak Purkait, Tat-Jun Chin, Ian Reid
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 FineNet fine-tuning after CleanNet-SPT (ablation) | CleanNet-SPT · Synthetic view-graphs | mean angular error (deg): 2.93 -> 1.45 | Improved |
| add CleanNet view-graph cleaning before SPT and FineNet (ablation) | SPT-FineNet · Synthetic view-graphs | mean angular error (deg): 3.00 -> 1.45 | Improved |
| run FineNet for two steps instead of one (ablation) | NeuRoRA · Synthetic view-graphs | mean angular error (deg): 1.45 -> 1.30 | Improved |
| initialize Chatterjee with CleanNet-SPT output | Chatterjee · Synthetic view-graphs | mean angular error (deg): 2.17 -> 2.11 | No clear effect |
| initialize Weiszfeld with CleanNet-SPT output | Weiszfeld · Synthetic view-graphs | mean angular error (deg): 3.35 -> 1.74 | Improved |
| initialize Weiszfeld with NeuRoRA output instead of CleanNet-SPT | Weiszfeld · Synthetic view-graphs | mean angular error (deg): 1.74 -> 1.73 | No clear effect |
| initialize Wang with CleanNet-SPT output | Wang · Synthetic view-graphs | mean angular error (deg): 2.77 -> 2.77 | No clear effect |
| replace Chatterjee with NeuRoRA | Chatterjee · Synthetic view-graphs | cpu runtime: 5.38s -> 0.21s | Improved |
| train NeuRoRA on planar cameras and test on non-planar | NeuRoRA · Synthetic, planar generalization | mean angular error (deg): 0.9 -> 2.2 | Hurt |
| replace Chatterjee with NeuRoRA on SanFran | Chatterjee · 1DSfM SanFran | mean angular error (deg): 3.6 -> 17.6 | Hurt |
| replace Chatterjee with NeuRoRA on GendrmMarkt | Chatterjee · 1DSfM GendrmMarkt | mean angular error (deg): 37.6 -> 4.5 | Improved |
| add NeuRoRA over CleanNet-SPT (ablation) | CleanNet-SPT · ScanNet view-graphs | mean angular error (deg): 11.37 -> 11.02 | Improved |
Later papers that compared with this method
1 later accepted paper reports NeuRoRA as a baseline in their results tables; 1 of them report a better number on most of the results they share with it.
- RAGO: Recurrent Graph Optimizer for Multiple Rotation Averaging (CVPR 2022 · better on 2 of 3 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
- End-to-End Rotation Averaging With Multi-Source Propagation (CVPR 2021 · similarity 0.89)
- HARA: A Hierarchical Approach for Robust Rotation Averaging (CVPR 2022 · similarity 0.86)
- Certifiably Optimal Anisotropic Rotation Averaging (ICCV 2025 · similarity 0.86)
- RAGO: Recurrent Graph Optimizer for Multiple Rotation Averaging (CVPR 2022 · similarity 0.86)
- Robust Relative Rotation Averaging (TPAMI 2017 · similarity 0.85)
- EqNIO: Subequivariant Neural Inertial Odometry (ICLR 2025 · similarity 0.82)
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.
