Research map / Point clouds / Registration and matching
TurboReg: TurboClique for Robust and Efficient Point Cloud Registration
Shaocheng Yan, Pengcheng Shi, Zhenjun Zhao, Kaixin Wang, Kuang Cao, Ji Wu, Jiayuan Li
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 |
|---|---|---|---|
| use TurboReg (1K) instead of 3DMAC | 3DMatch+FCGF pipeline · 3DMatch+FPFH | registration recall and speed: 83.92 -> 83.92 RR; 0.31 -> 2.73 CPU FPS | Improved |
| use TurboReg (2K) instead of 3DMAC with FPFH on 3DLoMatch | TurboReg · 3DLoMatch+FPFH | registration recall: 40.88 -> 40.99 | Improved |
| use TurboReg (1K) with Predator instead of 3DMAC | TurboReg · 3DLoMatch+Predator | registration recall: 70.90 -> 73.07 | Improved |
| add very tight compatibility threshold tau=0.0005 m (versus 0.012 m) (ablation) | TurboReg · 3DMatch+FPFH | registration recall: 84.10 -> 66.30 | Hurt |
| enlarge compatibility threshold to 0.5 m (ablation) | TurboReg · 3DMatch+FPFH | registration recall: 84.10 -> 74.43 | Hurt |
| replace undirected SC2 graph with directed O2Graph (ablation) | TurboReg · 3DMatch+FPFH | registration recall: 83.61 -> 84.10 (+0.49%) | Improved |
| use directed O2Graph instead of undirected SC2 graph (ablation) | TurboReg · 3DMatch+FPFH CPU time | processing time: 67.88 ms reduction | Improved |
| use directed O2Graph instead of undirected SC2 graph (ablation) | TurboReg · 3DMatch+FPFH GPU time | processing time: 0.69 ms difference | No clear effect |
| increase pivot number K1 from 10 to 2000 (ablation) | TurboReg · 3DMatch+FPFH | registration recall: 78.68 -> 84.10 | Improved |
| increase K1 beyond 250 (ablation) | TurboReg · 3DMatch+FCGF | registration recall: 93.28 -> 93.59 | No clear effect |
| increase TurboCliques per pivot K2 from 2 to 3 (ablation) | TurboReg · 3DMatch+FCGF | registration recall: 93.59 -> 93.53 | Hurt |
| decrease K2 from 2 to 1 (ablation) | TurboReg · 3DMatch+FPFH | registration recall: 84.10 -> 83.43 | Hurt |
Closest papers by meaning
- MHopReg: Efficient Hierarchical Multi-Hop Graph Search for Point Cloud Registration (CVPR 2026 · similarity 0.91)
- MAC: Maximal Cliques for 3D Registration (TPAMI 2024 · similarity 0.90)
- Efficient and Robust Point Cloud Registration via Heuristics-Guided Parameter Search (TPAMI 2024 · similarity 0.90)
- PointMC: Multi-instance Point Cloud Registration based on Maximal Cliques (ICML 2024 · similarity 0.90)
- Cross-PCR: A Robust Cross-Source Point Cloud Registration Framework (AAAI 2025 · similarity 0.89)
- A robust inlier identification algorithm for point cloud registration via $\mathbf{\ell_0}$-minimization (NeurIPS 2024 · 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.
