Research map / Point clouds / Point cloud classification
Learnable Skeleton-Aware 3D Point Cloud Sampling
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 skeleton-aware LFS prior sampling rather than the strongest reported baseline | Point cloud sampling with PointNet · ModelNet40 classification, ratio 64 | accuracy: 61.6 -> 82.9 | Helped |
| use skeleton-aware LFS prior sampling rather than the strongest reported baseline | Point cloud sampling with PointNet · ModelNet40 retrieval, ratio 64 | mAP: 59.5 -> 62.6 | Helped |
| use skeleton-aware LFS prior sampling rather than the strongest reported baseline | Point cloud sampling with PCN · ModelNet40 reconstruction, ratio 64 | CD: 4.31 -> 3.78 | Helped |
| add the initial LFS sampling weight (ablation) | Skeleton-aware sampling with PointNet · ModelNet40 classification, ratio 64 | accuracy: 80.3 -> 82.9 | Helped |
| use ground-truth skeletons from DPC instead of estimated skeletons (ablation) | Skeleton-aware sampling with PointNet · ModelNet40 classification, ratio 64 | accuracy: 82.8 -> 82.9 | No clear effect |
| use 6 LFS histogram bins instead of 1 bin (ablation) | Skeleton-aware sampling with PointNet · ModelNet40 classification, ratio 64 | accuracy: 80.3 -> 82.9 | Helped |
| vary the LFS histogram from 6 to 12 bins (ablation) | Skeleton-aware sampling with PointNet · ModelNet40 classification, ratio 8 | accuracy: 89.1 -> 89.1 | No clear effect |
| vary the LFS histogram from 6 to 18 bins (ablation) | Skeleton-aware sampling with PointNet · ModelNet40 classification, ratio 64 | accuracy: 82.9 -> 82.7 | No clear effect |
| increase skeletal spheres from 32 to 128 (ablation) | Skeleton estimation for PointNet sampling · ModelNet40 classification, ratio 64 | accuracy: 79.3 -> 82.9 | Helped |
| increase skeletal spheres from 128 to 256 (ablation) | Skeleton estimation for PointNet sampling · ModelNet40 classification, ratio 64 | accuracy: 82.9 -> 82.8 | No clear effect |
| anneal Gumbel-softmax temperature exponentially rather than keeping it constant (ablation) | Skeleton-aware sampling with PointNet · ModelNet40 classification, ratio 64 | accuracy: 81.1 -> 82.9 | Helped |
| use step temperature annealing rather than constant temperature (ablation) | Skeleton-aware sampling with PointNet · ModelNet40 classification, ratio 64 | accuracy: 81.1 -> 82.1 | Helped |
Closest papers by meaning
- SAMBLE: Shape-Specific Point Cloud Sampling for an Optimal Trade-Off Between Local Detail and Global Uniformity (CVPR 2025 · similarity 0.89)
- SPEAL: Skeletal Prior Embedded Attention Learning for Cross-Source Point Cloud Registration (AAAI 2024 · similarity 0.88)
- GCD-Sampling: A General Cross-scale Decoupled Sampling for Point Cloud (AAAI 2025 · similarity 0.87)
- SMART-PC: Skeletal Model Adaptation for Robust Test-Time Training in Point Clouds (ICML 2025 · similarity 0.87)
- Point-Cache: Test-time Dynamic and Hierarchical Cache for Robust and Generalizable Point Cloud Analysis (CVPR 2025 · similarity 0.87)
- LTA-PCS: Learnable Task-Agnostic Point Cloud Sampling (CVPR 2024 · similarity 0.87)
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.