Research map / Label-efficient and robust learning / Recognition and long-tail
End-to-End Rotation Averaging With Multi-Source Propagation
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 image node and correspondence edge features alongside relative rotations to AGF (ablation) | End-to-End Rotation Averaging · YFCC100 test sets | average mean angular error: 3.23 vs. 4.18; 3.23 vs. 3.98 | Helped |
| Use M=15 reference nodes instead of M=1 (ablation) | Multi-Source Propagation · YFCC100 test sets | average mean angular error: 4.65 -> 3.23 | Helped |
| Use M=15 reference nodes instead of M=5 (ablation) | Multi-Source Propagation · YFCC100 test sets | average mean angular error: 4.66 -> 3.23 | Helped |
| Use M=15 reference nodes instead of M=10 (ablation) | Multi-Source Propagation · YFCC100 test sets | average mean angular error: 4.1 -> 3.23 | Helped |
| Replace MSP with SPT-based initialization (ablation) | End-to-End Rotation Averaging · YFCC100 test sets | average mean angular error: 3.23 -> 9.37 | Hurt |
| Add iterative edge-weight refinement at test time (ablation) | End-to-End Rotation Averaging · YFCC100: colosseum exterior | mean angular error: NeuRoRA: 22.81; NeuRoRA + Refine: 3.06 | Helped |
| Add iterative edge-weight refinement at test time (ablation) | NeuRoRA · YFCC100: palace of versailles chapel | mean angular error: 2.98 -> 5.01 | Hurt |
| Add iterative edge-weight refinement at test time (ablation) | NeuRoRA · YFCC100: big ben 1 | mean angular error: 5.22 -> 9.01 | Hurt |
| Add image context and matched-keypoint geometry to edge reliability prediction (ablation) | Appearance-Geometry Fusion · YFCC100 test sets | average mean angular error: 3.23 vs. 4.18 | Helped |
| Add relative rotations to AGF edge features (ablation) | Appearance-Geometry Fusion · YFCC100 test sets | average mean angular error: 3.23 vs. 4.18 | Helped |
| Increase the number of MSP reference nodes from M=1 to M=10 (ablation) | Multi-Source Propagation · YFCC100 test sets | average mean angular error: 4.65 -> 4.1 | Helped |
Closest papers by meaning
- HARA: A Hierarchical Approach for Robust Rotation Averaging (CVPR 2022 · similarity 0.87)
- RAGO: Recurrent Graph Optimizer for Multiple Rotation Averaging (CVPR 2022 · similarity 0.85)
- Hybrid Rotation Averaging: A Fast and Robust Rotation Averaging Approach (CVPR 2021 · similarity 0.84)
- Rotation Averaging and Strong Duality (CVPR 2018 · similarity 0.83)
- Learning to Filter Outlier Edges in Global SfM (CVPR 2025 · similarity 0.82)
- Robust Angular Synchronization via Directed Graph Neural Networks (ICLR 2024 · similarity 0.81)
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.