Research map / Point clouds / Registration and matching
Iterative Distance-Aware Similarity Matrix Convolution with Mutual-Supervised Point Elimination for Efficient Point Cloud Registration
Jiahao Li, Changhao Zhang, Ziyao Xu, Hangning Zhou, Chi Zhang
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 distance-aware similarity matrix convolution to the baseline (ablation) | BS (inner-product matching) · ModelNet40 unseen categories | RMSE(R) degrees: 7.77 -> 5.08 | Improved |
| add hard point elimination (ablation) | BS+SM · ModelNet40 unseen categories | RMSE(R) degrees: 5.08 -> 4.31 | Improved |
| add hybrid point elimination (ablation) | BS+HA+SM · ModelNet40 unseen categories | RMSE(R) degrees: 4.31 -> 3.42 | Improved |
| add hybrid point elimination (ablation) | BS+HA+SM · ModelNet40 unseen categories | MAE(R) degrees: 2.89 -> 0.93 | Improved |
| add hard point elimination (ablation) | BS+SM · ModelNet40 unseen categories | RMSE(t): 0.056 -> 0.029 | Improved |
| use learned GNN features instead of FPFH (ablation) | IDAM · ModelNet40 unseen shapes | RMSE(R) degrees: 2.46 -> 2.95 | Hurt |
| use GNN features instead of FPFH under Gaussian noise | IDAM · ModelNet40 unseen shapes with Gaussian noise | RMSE(R) degrees: 14.21 -> 3.72 | Improved |
| use GNN features instead of FPFH | IDAM · ModelNet40 unseen categories | RMSE(R) degrees: 3.04 -> 3.42 | Hurt |
| replace PRNet with IDAM (FPFH) on unseen shapes | FPFH+IDAM vs PRNet · ModelNet40 unseen shapes | MAE(R) degrees: 1.45 -> 0.56 | Improved |
| replace PRNet with GNN+IDAM at 4096 points | GNN+IDAM vs PRNet · Speed, 4096 points | seconds per frame: 0.312 -> 0.041 | Improved |
| use FPFH+IDAM instead of RANSAC or FGR with FPFH | FPFH+IDAM · Speed | seconds per frame: 2x to 5x faster | 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.
- PRNet — PRNet: Self-Supervised Learning for Partial-to-Partial Registration (NeurIPS 2019) · better on 6 of 6 shared results
- PointNetLK — PointNetLK: Robust & Efficient Point Cloud Registration Using PointNet (CVPR 2019) · better on 3 of 3 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
- OMNet: Learning Overlapping Mask for Partial-to-Partial Point Cloud Registration (ICCV 2021 · similarity 0.91)
- End-to-End Learning the Partial Permutation Matrix for Robust 3D Point Cloud Registration (AAAI 2022 · similarity 0.90)
- 3D Focusing-and-Matching Network for Multi-Instance Point Cloud Registration (NeurIPS 2024 · similarity 0.90)
- Implicit Correspondence Learning for Image-to-Point Cloud Registration (CVPR 2025 · similarity 0.88)
- PointMC: Multi-instance Point Cloud Registration based on Maximal Cliques (ICML 2024 · similarity 0.88)
- Partial Point Cloud Registration with Multi-view 2D Image Learning (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.
