Research map / Point clouds / Shape and surface completion
Learning and Matching Multi-View Descriptors for Registration of Point Clouds
Lei Zhou, Siyu Zhu, Zixin Luo, Tianwei Shen, Runze Zhang, Mingmin Zhen, Tian Fang, Long Quan
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 Fuseption-based FRN fusion instead of view pooling (ablation) | MVDesc · HPatches patch verification | mAP: 0.883 -> 0.921 | Improved |
| use Fuseption-based FRN fusion instead of view pooling (ablation) | MVDesc · HPatches patch retrieval | mAP: 0.456 -> 0.530 | Improved |
| use Fuseption-based FRN fusion instead of view pooling (ablation) | MVDesc · HPatches image matching | mAP: 0.312 -> 0.325 | Improved |
| use 3 views of a patch instead of a single view | MVDesc vs L2-Net · HPatches patch verification | mAP: 0.792 -> 0.921 | Improved |
| increase descriptor dimension (ablation) | MVDesc · HPatches image matching | mAP | Mixed |
| use MVDesc-128 instead of MVDesc-32 | MVDesc · TUM descriptor matching | precision: 0.664 -> 0.695 | Improved |
| use MVDesc-32 instead of CGF | descriptor matching · TUM descriptor matching | precision: 0.447 -> 0.664 | Improved |
| add RMBP outlier rejection to CGF (ablation) | CGF · ScanNet fragment registration | recall: 0.350 -> 0.419 | Improved |
| add RMBP outlier rejection to 3DMatch (ablation) | 3DMatch · ScanNet fragment registration | recall: 0.185 -> 0.145 | Hurt |
| add RMBP outlier rejection to MVDesc-128 (ablation) | MVDesc-128 · ScanNet fragment registration | precision: 0.910 -> 0.906 | Hurt |
| add RMBP outlier rejection to FPFH (ablation) | FPFH · ScanNet fragment registration | recall: 0.119 -> 0.272 | Improved |
| use RMBP instead of RANSAC, SMG or LPM at low inlier ratio | outlier rejection · TUM, ScanNet, EPFL point cloud pairs | valid registration | Improved |
Closest papers by meaning
- Learning general and distinctive 3D local deep descriptors for point cloud registration (TPAMI 2022 · similarity 0.90)
- End-to-End Learning Local Multi-View Descriptors for 3D Point Clouds (CVPR 2020 · similarity 0.90)
- 2D3D-MATR: 2D-3D Matching Transformer for Detection-Free Registration Between Images and Point Clouds (ICCV 2023 · similarity 0.89)
- Learning Multiview 3D Point Cloud Registration (CVPR 2020 · similarity 0.89)
- DV-Matcher: Deformation-based Non-rigid Point Cloud Matching Guided by Pre-trained Visual Features (CVPR 2025 · similarity 0.88)
- RIGA: Rotation-Invariant and Globally-Aware Descriptors for Point Cloud Registration (TPAMI 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.
