Research map / Point clouds / Registration and matching
Correspondence-Free SE(3) Point Cloud Registration in RKHS via Unsupervised Equivariant Learning
Ray Zhang, Zheming Zhou, Min Sun, Omid Ghasemalizadeh, Cheng-Hao Kuo, Ryan M. Eustice, Maani Ghaffari Jadidi, Arnie Sen
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 tanh kernel on steerable features to the RBF kernel (ablation) | EquivAlign · ModelNet40, init angle <45 | rotation error (deg): 81.97 -> 0.29 | Improved |
| add curriculum learning on initial angles (1 to 45 degrees) (ablation) | EquivAlign unsupervised training · ModelNet40, init angle <45 | rotation error mean (deg): 2.73 -> 0.29 | Improved |
| use larger initial kernel lengthscale for larger initial angles (ablation) | EquivAlign · ModelNet40 | rotation error | Improved |
| use EquivAlign instead of ICP under 20% outliers at 90 degrees | point cloud registration · ModelNet40, init angle <90, sigma=0.01, gamma=20% | rotation error: 38.00 -> 4.12 | Improved |
| use EquivAlign instead of GeoTransformer under noise-free 45 degrees | point cloud registration · ModelNet40, init angle <45, sigma=0.01, gamma=20% | rotation error: 0.92 -> 1.96 | Hurt |
| use EquivAlign instead of FPFH-based methods without noise at 90 degrees | point cloud registration · ModelNet40, init angle <90, no noise | rotation error: 0.14 -> 1.14 | Hurt |
| use EquivAlign with random cropping at 45 degrees | point cloud registration · ModelNet40, init angle <45, crop 20% | rotation error: 1.39 -> 19.20 | Hurt |
| use EquivAlign unsupervised instead of supervised GeoTransformer on real data | point cloud registration · ETH3D | rotation error mean (deg): 2.23 -> 0.53 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- GeoTransformer — GeoTransformer: Fast and Robust Point Cloud Registration With Geometric Transformer (TPAMI 2023) · better on 2 of 4 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
- Equi-GSPR: Equivariant SE(3) Graph Network Model for Sparse Point Cloud Registration (ECCV 2024 · similarity 0.89)
- Generalized-CVO: Fast and Correspondence-Free Local Point Cloud Registration with Second Order Riemannian Optimization (CVPR 2026 · similarity 0.89)
- SHReg: Strictly Rotation-Equivariant Point Cloud Registration via Spherical Harmonics (ECCV 2026 · similarity 0.88)
- Unsupervised RGB-D Point Cloud Registration for Scenes with Low Overlap and Photometric Inconsistency (ICCV 2025 · similarity 0.88)
- Extend Your Own Correspondences: Unsupervised Distant Point Cloud Registration by Progressive Distance Extension (CVPR 2024 · similarity 0.88)
- A robust inlier identification algorithm for point cloud registration via $\mathbf{\ell_0}$-minimization (NeurIPS 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.
