Research map / Point clouds / Registration and matching
3D Registration With Maximal Cliques
Read the paperCheck 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 |
|---|---|---|---|
| Use maximal cliques rather than only maximum cliques (ablation) | MAC · 3DMatch and 3DLoMatch | registration recall: 9.8%; 5.55% higher on 3DMatch, and 8.03%; 10.45% higher on 3DLoMatch | Helped |
| Add node-guided clique selection (ablation) | MAC · 3DMatch and 3DLoMatch | registration recall: 0.37%; 0.5% on 3DMatch, and 0.23%; 0.73% on 3DLoMatch | Helped |
| Add geometric consistency filtering before MAC (ablation) | MAC · 3DMatch and 3DLoMatch | registration recall: 83.86 -> 77.02; 93.41 -> 91.68 | Hurt |
| Use SOG rather than FOG (ablation) | MAC · 3DMatch and 3DLoMatch | registration recall: 1.6%; 0.06% higher on 3DMatch, and 0.12%; 0.56% higher on 3DLoMatch | Helped |
| Add normal-consistency filtering (ablation) | MAC · 3DMatch and 3DLoMatch | registration recall: 83.86 -> 82.44; 93.41 -> 93.28 | Hurt |
| Use weighted SVD instead of instance-equal SVD (ablation) | MAC · 3DMatch and 3DLoMatch | registration recall: 83.86 -> 83.67; 93.41 -> 93.16 | Hurt |
| Use MAE rather than inlier count for hypothesis evaluation (ablation) | MAC · 3DMatch and 3DLoMatch | registration recall: 0.24%; 0.31% higher on 3DMatch, and 1.74%; 0.05% higher on 3DLoMatch | Mixed |
| Add MAC to learned registration methods | FCGF, SpinNet, Predator, CoFiNet, and GeoTransformer · 3DMatch and 3DLoMatch | registration recall: 3.7–14.2 percentage points on 3DMatch and 3.1–17.1 percentage points on 3DLoMatch, as printed across tested settings | Helped |
Closest papers by meaning
- PointMC: Multi-instance Point Cloud Registration based on Maximal Cliques (ICML 2024 · similarity 0.91)
- FastMAC: Stochastic Spectral Sampling of Correspondence Graph (CVPR 2024 · similarity 0.88)
- 3D Focusing-and-Matching Network for Multi-Instance Point Cloud Registration (NeurIPS 2024 · similarity 0.87)
- HeMoRa: Unsupervised Heuristic Consensus Sampling for Robust Point Cloud Registration (CVPR 2025 · similarity 0.85)
- Partial Point Cloud Registration with Multi-view 2D Image Learning (AAAI 2025 · similarity 0.85)
- Cross-PCR: A Robust Cross-Source Point Cloud Registration Framework (AAAI 2025 · similarity 0.85)
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.