Research map / Point clouds / Registration and matching
RSKDD-Net: Random Sample-based Keypoint Detector and Descriptor
Fan Lu, Guang Chen, Yinlong Liu, Zhongnan Qu, Alois Knoll
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 RSKDD-Net (random sampling + learned keypoints/descriptors) instead of USIP | keypoint detector and descriptor · KITTI registration | inlier ratio: 0.243 -> 0.586 | Improved |
| use RSKDD-Net instead of USIP | keypoint detector and descriptor · KITTI registration | RANSAC iterations: 468 -> 32 | Improved |
| use RSKDD-Net instead of USIP | keypoint detector and descriptor · KITTI registration | RRE (deg): 0.35 -> 0.50 | Hurt |
| use RSKDD-Net instead of USIP | keypoint detector and descriptor · KITTI registration | RTE (m): 0.10 -> 0.09 | Improved |
| use random sampling instead of FPS/per-point saliency | keypoint detection · KITTI, 16384 points, 512 keypoints | time (ms): 378.5 -> 10.1 | Improved |
| add random dilation cluster (versus plain kNN cluster) (ablation) | RSKDD-Net · KITTI | repeatability and precision | Improved |
| use random dilation instead of dilated point convolution (DPC) (ablation) | RSKDD-Net · KITTI | precision | No clear effect |
| add attentive feature map to descriptor network (ablation) | RSKDD-Net · KITTI | precision: +0.1 precision | Improved |
| replace triplet loss with proposed soft-assignment matching loss (ablation) | RSKDD-Net · KITTI | precision: about 2x precision | Improved |
| increase matching-loss temperature t to 0.5 (ablation) | RSKDD-Net · KITTI | precision | Hurt |
| add keypoint saliency weight to matching loss (ablation) | RSKDD-Net · KITTI | precision | 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.
- USIP — USIP: Unsupervised Stable Interest Point Detection From 3D Point Clouds (ICCV 2019) · better on 8 of 11 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
- Sparse-to-Dense Matching Network for Large-Scale LiDAR Point Cloud Registration (TPAMI 2023 · similarity 0.88)
- HRegNet: A Hierarchical Network for Large-Scale Outdoor LiDAR Point Cloud Registration (ICCV 2021 · similarity 0.88)
- Density-invariant Features for Distant Point Cloud Registration (ICCV 2023 · similarity 0.88)
- Not All Points Are Equal: Learning Highly Efficient Point-Based Detectors for 3D LiDAR Point Clouds (CVPR 2022 · similarity 0.88)
- HRegNet: A Hierarchical Network for Efficient and Accurate Outdoor LiDAR Point Cloud Registration (TPAMI 2023 · similarity 0.88)
- Mining and Transferring Feature-Geometry Coherence for Unsupervised Point Cloud Registration (NeurIPS 2024 · similarity 0.86)
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.
