Research map / Point clouds / Point cloud classification
A Closer Look at Few-Shot 3D Point Cloud Classification
Chuangguan Ye, Hongyuan Zhu, Bo Zhang, Tao Chen
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 SPF module (ablation) | Prototypical Net (DGCNN) · ModelNet40-FS 5w-1s | accuracy (%): 69.95 -> 71.89 | Improved |
| add SCI+ module (ablation) | Prototypical Net (DGCNN) · ModelNet40-FS 5w-1s | accuracy (%): 69.95 -> 72.48 | Improved |
| add CIF+ module (ablation) | Prototypical Net (DGCNN) · ModelNet40-FS 5w-1s | accuracy (%): 69.95 -> 78.54 | Improved |
| add all three modules SPF+SCI+CIF+ (ablation) | Prototypical Net (DGCNN) · ModelNet40-FS 5w-1s | accuracy (%): 69.95 -> 81.19 | Improved |
| add all three modules SPF+SCI+CIF+ (ablation) | Prototypical Net (DGCNN) · ScanObjectNN-FS 5w-1s | accuracy (%): 46.28 -> 56.24 | Improved |
| add SPF module (ablation) | Prototypical Net (DGCNN) · ScanObjectNN-FS 5w-5s | accuracy (%): 63.94 -> 67.56 | Improved |
| use Prototypical Net on PointNet instead of vanilla pretrain-finetune | PointNet · ScanObjectNN-FS 5w-1s | accuracy (%): 40.00 -> 44.75 | Improved |
| use Prototypical Net on RSCNN instead of vanilla few-shot classification | RSCNN · ScanObjectNN-FS 5w-1s | accuracy (%): 31.88 -> 48.58 | Improved |
| use metric-based instead of optimization-based FSL on PointNet | Prototypical Net vs MAML · ModelNet40-FS 5w-1s | accuracy (%): 57.58 -> 65.31 | Improved |
| use DGCNN instead of PointNet as backbone for Prototypical Net | Prototypical Net · ModelNet40-FS 5w-1s | accuracy (%): 65.31 -> 69.95 | Improved |
| increase N in N-way classification from 10 to 20 | PCIA / Prototypical Net · ShapeNet70-FS 1-shot | accuracy (%): 56.08 -> 44.03 | Hurt |
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.
- Point-BERT — Point-BERT: Pre-Training 3D Point Cloud Transformers With Masked Point Modeling (CVPR 2022) · 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
- Hyperbolic Uncertainty-Aware Few-Shot Incremental Point Cloud Segmentation (CVPR 2025 · similarity 0.88)
- Rethinking Few-shot 3D Point Cloud Semantic Segmentation (CVPR 2024 · similarity 0.88)
- Generalized Few-shot 3D Point Cloud Segmentation with Vision-Language Model (CVPR 2025 · similarity 0.88)
- Geometric Feature Embedding for Effective 3D Few-Shot Class Incremental Learning (ICML 2025 · similarity 0.87)
- Multimodality Helps Few-shot 3D Point Cloud Semantic Segmentation (ICLR 2025 · similarity 0.87)
- Multi-View Representation is What You Need for Point-Cloud Pre-Training (ICLR 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.
