Research map / Point clouds / Segmentation and detection
Prototypical VoteNet for Few-Shot 3D Point Cloud Object Detection
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 |
|---|---|---|---|
| add PVM to baseline (ablation) | Baseline · FS-ScanNet Split 1 3-shot | AP25: 22.64 -> 27.43 | Helped |
| add PVM to baseline (ablation) | Baseline · FS-ScanNet Split 1 3-shot | AP50: 9.04 -> 13.63 | Helped |
| add PHM to baseline (ablation) | Baseline · FS-ScanNet Split 1 3-shot | AP25: 22.64 -> 28.76 | Helped |
| add PHM to baseline (ablation) | Baseline · FS-ScanNet Split 1 3-shot | AP50: 9.04 -> 14.04 | Helped |
| combine PVM and PHM (ablation) | Baseline · FS-ScanNet Split 1 5-shot | AP25: 24.93 -> 32.25 | Helped |
| replace geometric prototypes with self-learning prototypes in PVM (ablation) | Prototypical VoteNet · FS-ScanNet Split 1 3-shot | AP25: 31.25 -> 28.34 | Hurt |
| replace class prototypes with self-learning prototypes in PHM (ablation) | Prototypical VoteNet · FS-ScanNet Split 1 3-shot | AP25: 31.25 -> 27.45 | Hurt |
| increase memory bank size from 30 to 120 prototypes (ablation) | Prototypical VoteNet · FS-ScanNet Split 1 3-shot | AP25: 29.98 -> 31.25 | Helped |
| increase memory bank size from 120 to 150 prototypes (ablation) | Prototypical VoteNet · FS-ScanNet Split 1 3-shot | AP25: 31.25 -> 31.01 | No clear effect |
| momentum coefficient 0.2 vs 0.999 (ablation) | Prototypical VoteNet · FS-ScanNet Split 1 3-shot | AP25: 31.01 -> 29.50 | Hurt |
| add pretraining stage to VoteNet+TFA (ablation) | VoteNet+TFA · FS-ScanNet Split 1 3-shot | AP25: 8.07 -> 10.37 | Helped |
| also train VoteNet on novel classes in first TFA stage (TFA*) (ablation) | VoteNet+TFA · FS-ScanNet Split 1 5-shot | AP25: 16.36 -> 26.02 | Helped |
Closest papers by meaning
- ImVoteNet: Boosting 3D Object Detection in Point Clouds With Image Votes (CVPR 2020 · similarity 0.90)
- Prototypical Variational Autoencoder for 3D Few-shot Object Detection (NeurIPS 2023 · similarity 0.89)
- Back-Tracing Representative Points for Voting-Based 3D Object Detection in Point Clouds (CVPR 2021 · similarity 0.89)
- X-3D: Explicit 3D Structure Modeling for Point Cloud Recognition (CVPR 2024 · similarity 0.86)
- Rethinking Few-shot 3D Point Cloud Semantic Segmentation (CVPR 2024 · similarity 0.86)
- Hyperbolic Uncertainty-Aware Few-Shot Incremental Point Cloud Segmentation (CVPR 2025 · 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.