Research map / Point clouds / Registration and matching
Cross3DReg: A Real-world Cross-source Point Cloud Registration Benchmark
Zongyi Xu, Z. Lang, Yilong Chen, Shanshan Zhao, Xiaoshui Huang, Sean Bin Yang, Yan Zhang, Qianni Zhang, Xinbo Gao
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 overlapping mask prediction (OMP) with unaligned images (ablation) | geometric self-attention matcher · Cross3DReg (LGR) | RRE (deg): 18.184 -> 8.371 | Improved |
| add overlapping mask prediction (OMP) (ablation) | geometric self-attention matcher · Cross3DReg (LGR) | RR (%): 81.7 -> 86.7 | Improved |
| replace geometric self-attention with visual-geometric attention matching (VGAM) (ablation) | OMP with geometric self-attention · Cross3DReg (LGR) | RRE (deg): 8.371 -> 6.683 | Improved |
| replace vanilla self-attention with visual-geometric attention matching (VGAM) (ablation) | OMP with vanilla self-attention · Cross3DReg (LGR) | RTE (m): 1.242 -> 1.010 | Improved |
| use geometric instead of vanilla self-attention with OMP (ablation) | OMP · Cross3DReg (LGR) | RTE (m): 1.242 -> 1.287 | Hurt |
| use Cross3DReg instead of GeoTrans | GeoTrans · Cross3DReg (LGR) | RR (%): 81.7 -> 87.1 | Improved |
| use Cross3DReg instead of GeoTrans | GeoTrans · Cross3DReg (LGR) | RRE (deg): 18.15 -> 6.68 | Improved |
| use RANSAC-50K estimator with Cross3DReg | Cross3DReg vs Geotrans · Cross3DReg (RANSAC-50K) | RR (%): 72.2 -> 78.6 | Improved |
| use weighted SVD estimator with Cross3DReg | Cross3DReg vs Geotrans · Cross3DReg (Weighted SVD) | RR (%): 62.9 -> 57.8 | Hurt |
| use same-source methods (FCGF, Predator, CoFiNet, Omnet) on cross-source data | same-source registration · Cross3DReg (LGR) | RR (%): 0.0 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- CoFiNet — CoFiNet: Reliable Coarse-to-fine Correspondences for Robust PointCloud Registration (NeurIPS 2021) · better on 1 of 1 shared result
- Omnet — OMNet: Learning Overlapping Mask for Partial-to-Partial Point Cloud Registration (ICCV 2021) · better on 1 of 1 shared result
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
- Cross-PCR: A Robust Cross-Source Point Cloud Registration Framework (AAAI 2025 · similarity 0.93)
- FreeReg: Image-to-Point Cloud Registration Leveraging Pretrained Diffusion Models and Monocular Depth Estimators (ICLR 2024 · similarity 0.90)
- C-GenReg: Training-Free 3D Point Cloud Registration by Multi-View-Consistent Geometry-to-Image Generation with Probabilistic Modalities Fusion (CVPR 2026 · similarity 0.90)
- RoReg: Pairwise Point Cloud Registration With Oriented Descriptors and Local Rotations (TPAMI 2023 · similarity 0.89)
- $\pi^2$: A Simple Framework for 2D-to-3D Registration (NeurIPS 2026 · similarity 0.89)
- HybridReg: Robust 3D Point Cloud Registration with Hybrid Motions (AAAI 2025 · similarity 0.88)
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.
