Research map / Point clouds / Registration and matching
SHReg: Strictly Rotation-Equivariant Point Cloud Registration via Spherical Harmonics
Chongjian Wang, Junjie Gao
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 SH/irreps equivariant backbone instead of rotation-sensitive baseline (ablation) | SHReg · 3DLoMatch | RR: 78.2 -> 82.4 | Improved |
| add Clebsch-Gordan projection (ablation) | SHReg backbone · 3DLoMatch | RR: 79.3 -> 82.4 | Improved |
| add gated nonlinearity (ablation) | SHReg backbone · 3DLoMatch | RR: 80.1 -> 82.4 | Improved |
| add invariant descriptor readout (ablation) | SHReg backbone · 3DLoMatch | RR: 75.5 -> 82.4 | Improved |
| add invariant descriptor readout (ablation) | SHReg backbone · 3DMatch | IR: 68.0 -> 78.6 | Improved |
| use single-correspondence hypothesis proposer instead of RANSAC (ablation) | SHReg · 3DLoMatch | RR: 79.5 -> 82.4 | Improved |
| use single-correspondence hypothesis proposer instead of LGR (ablation) | SHReg · 3DLoMatch | RR: 81.0 -> 82.4 | Improved |
| use single-correspondence hypothesis proposer instead of RANSAC (ablation) | SHReg · 3DMatch | RR: 94.0 -> 95.4 | Improved |
| apply full-range rotation to 3DLoMatch | SHReg · Rotated 3DLoMatch | TR: 83.0 -> 83.3 | No clear effect |
| apply full-range rotation to 3DLoMatch | FCGF · Rotated 3DLoMatch | TR: 39.6 -> 24.5 | Hurt |
| apply full-range rotation to 3DLoMatch | GeoTrans · Rotated 3DLoMatch | TR: 75.4 -> 72.6 | Hurt |
| evaluate on KITTI without fine-tuning (trained on 3DMatch) | SHReg · KITTI Odometry | TE: 4.9 -> 4.7 | Improved |
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.
- PEAL — PEAL: Prior-Embedded Explicit Attention Learning for Low-Overlap Point Cloud Registration (CVPR 2023) · better on 3 of 3 shared results
- BUFFER — BUFFER: Balancing Accuracy, Efficiency, and Generalizability in Point Cloud Registration (CVPR 2023) · 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
- RoReg: Pairwise Point Cloud Registration With Oriented Descriptors and Local Rotations (TPAMI 2023 · similarity 0.91)
- PARE-Net: Position-Aware Rotation-Equivariant Networks for Robust Point Cloud Registration (ECCV 2024 · similarity 0.91)
- Rethinking Rotation Invariance with Point Cloud Registration (AAAI 2023 · similarity 0.89)
- TurboReg: TurboClique for Robust and Efficient Point Cloud Registration (ICCV 2025 · similarity 0.89)
- A robust inlier identification algorithm for point cloud registration via $\mathbf{\ell_0}$-minimization (NeurIPS 2024 · similarity 0.87)
- FreeReg: Image-to-Point Cloud Registration Leveraging Pretrained Diffusion Models and Monocular Depth Estimators (ICLR 2024 · similarity 0.87)
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.
