Research map / Point clouds / Point cloud classification
Learning general and distinctive 3D local deep descriptors for point cloud registration
Fabio Poiesi, Davide Boscaini
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 QNet spatial transformer (ablation) | GeDi · 3DMatch ablation (7-scenes subset) | feature-matching recall: .917 -> .932 | Improved |
| add LRF canonicalisation (ablation) | GeDi · 3DMatch ablation (7-scenes subset) | feature-matching recall: .787 -> .932 | Improved |
| add LRF canonicalisation (ablation) | GeDi · 3DMatchR ablation | feature-matching recall: .019 -> .931 | Improved |
| increase descriptor dimension from 16 to 128 (ablation) | GeDi · 3DMatch ablation (7-scenes subset) | feature-matching recall: .902 -> .937 | Improved |
| increase input points from 256 to 2048 (ablation) | GeDi · 3DMatch ablation (7-scenes subset) | feature-matching recall: .903 -> .937 | Improved |
| reduce patch radius from .6 to .2 m (ablation) | GeDi · 3DMatch ablation (7-scenes subset) | feature-matching recall: .932 -> .741 | Hurt |
| increase patch radius from .6 to 1.8 m (ablation) | GeDi · 3DMatch ablation (7-scenes subset) | feature-matching recall: .932 -> .315 | Hurt |
| replace QNet with TNet plus regularisation loss (ablation) | GeDi · 3DMatch ablation (7-scenes subset) | feature-matching recall: .932 -> .932 | No clear effect |
| replace QNet with TNet plus SVD orthogonalisation (ablation) | GeDi · 3DMatch ablation (7-scenes subset) | feature-matching recall: .932 -> .934 | No clear effect |
| apply QNet rotation after LRF (ablation) | GeDi · ablation pairs (1.4M) | median RRE: 0.909 -> 0.497 | Improved |
| use GeDi instead of DIP | DIP · 3DMatch->ETH | average FMR: .928 -> .982 | Improved |
| use GeDi instead of SpinNet with 250 sampled points | SpinNet · 3DMatch | feature-matching recall: .943 -> .973 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- SpinNet — SpinNet: Learning a General Surface Descriptor for 3D Point Cloud Registration (CVPR 2021) · better on 10 of 10 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
- 3D Point Cloud Registration for Localization Using a Deep Neural Network Auto-Encoder (CVPR 2017 · similarity 0.92)
- SpinNet: Learning a General Surface Descriptor for 3D Point Cloud Registration (CVPR 2021 · similarity 0.92)
- You Only Train Once: Learning General and Distinctive 3D Local Descriptors (TPAMI 2022 · similarity 0.92)
- RIGA: Rotation-Invariant and Globally-Aware Descriptors for Point Cloud Registration (TPAMI 2024 · similarity 0.91)
- DH3D: Deep Hierarchical 3D Descriptors for Robust Large-Scale 6DoF Relocalization (ECCV 2020 · similarity 0.91)
- The Perfect Match: 3D Point Cloud Matching With Smoothed Densities (CVPR 2019 · similarity 0.91)
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.
