Research map / Detection and segmentation / Open-set and few-shot detection
Prototypical Variational Autoencoder for 3D Few-shot 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 GP-VAE (ablation) | VoteNet baseline · FS-ScanNet split-1 3-shot | AP50: 9.57 -> 15.18 | Helped |
| add CP-VAE (ablation) | VoteNet baseline · FS-ScanNet split-1 3-shot | AP50: 9.57 -> 18.88 | Helped |
| add CP-VAE on top of GP-VAE (ablation) | P-VAE · FS-ScanNet split-1 3-shot | AP50: 15.18 -> 19.37 | Helped |
| add CP-VAE on top of GP-VAE (ablation) | P-VAE · FS-ScanNet split-1 5-shot | AP25: 33.48 -> 32.84 | Mixed |
| use CP-VAE without GP-VAE (ablation) | P-VAE · FS-ScanNet split-1 3-shot | AP25: 31.60 -> 27.11 | Hurt |
| replace proto learning with AE in GP-VAE (ablation) | GP-VAE · FS-ScanNet split-1 3-shot | AP50: 13.57 -> 14.48 | Helped |
| replace AE with VAE in GP-VAE (ablation) | GP-VAE · FS-ScanNet split-1 5-shot | AP50: 19.16 -> 22.18 | Helped |
| replace AE with VAE in CP-VAE (ablation) | CP-VAE · FS-ScanNet split-1 3-shot | AP50: 17.98 -> 18.88 | Helped |
| replace proto learning with AE in CP-VAE (ablation) | CP-VAE · FS-ScanNet split-1 5-shot | AP50: 19.16 -> 21.45 | Helped |
| increase number of geometric prototypes Ngeo from 16 to 128 (ablation) | GP-VAE · FS-ScanNet split-1 5-shot | AP50: 6.48 -> 22.39 | Helped |
| increase Ngeo from 128 to 256 (ablation) | GP-VAE · FS-ScanNet split-1 5-shot | AP50: 22.39 -> 20.30 | Hurt |
| use Prototypical VoteNet instead of VoteNet baseline for base classes (ablation) | VoteNet · FS-ScanNet-1 base classes | AP50: 32.60 -> 28.20 | Hurt |
Closest papers by meaning
- ImOV3D: Learning Open Vocabulary Point Clouds 3D Object Detection from Only 2D Images (NeurIPS 2024 · similarity 0.85)
- Dora: Sampling and Benchmarking for 3D Shape Variational Auto-Encoders (CVPR 2025 · similarity 0.85)
- Towards Compact 3D Representations via Point Feature Enhancement Masked Autoencoders (AAAI 2024 · similarity 0.85)
- Hyperbolic Uncertainty-Aware Few-Shot Incremental Point Cloud Segmentation (CVPR 2025 · similarity 0.84)
- Fine-Grained Prototypes Distillation for Few-Shot Object Detection (AAAI 2024 · similarity 0.84)
- Generalized Few-shot 3D Point Cloud Segmentation with Vision-Language Model (CVPR 2025 · similarity 0.84)
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.