Research map / Point clouds / Segmentation and detection
From Coarse to Fine: Deep Prototype Refinement Network for Few-Shot Point Cloud Semantic Segmentation
Changshuo Wang, Weijun Li, Shuting He, Xiang Fang, Xingyu Gao, Zhonghang Liu, Prayag Tiwari, Dimitrios Kanoulas
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 spatial cross-correlation features (ablation) | DPR module · S3DIS 2-way 1-shot S0 | mIoU: 47.31 -> 80.76 | Improved |
| add common subspace features (ablation) | DPR module · S3DIS 2-way 1-shot S0 | mIoU: 78.56 -> 80.76 | Improved |
| add dual-subspace decomposition (ablation) | DPR module · S3DIS 2-way 1-shot S0 | mIoU: 78.99 -> 80.76 | Improved |
| add distinctive subspace features (ablation) | DPR module · S3DIS 2-way 1-shot S0 | mIoU: 79.15 -> 80.76 | Improved |
| add channel correlation loss (ablation) | DPR-Net · S3DIS 2-way 1-shot S0 | mIoU: 78.62 -> 78.95 | Improved |
| add inter-class diversity loss (ablation) | DPR-Net · S3DIS 2-way 1-shot S0 | mIoU: 78.95 -> 79.78 | Improved |
| add inter-generation diversity loss (ablation) | DPR-Net · S3DIS 2-way 1-shot S0 | mIoU: 79.78 -> 80.76 | Improved |
| increase refinement iterations from 1 to 3 (ablation) | DPR-Net · S3DIS 2-way 1-shot S0 | mIoU: 45.83 -> 80.76 | Improved |
| increase refinement iterations from 3 to 5 (ablation) | DPR-Net · S3DIS 2-way 1-shot S0 | mIoU: 80.76 -> 78.75 | Hurt |
| use confidence-based MoPE instead of simple majority voting (ablation) | DPR-Net · S3DIS 2-way 1-shot S0 | mIoU: 76.97 -> 80.76 | Improved |
| use confidence-based MoPE instead of learnable weights (ablation) | DPR-Net · S3DIS 2-way 1-shot S0 | mIoU: 79.43 -> 80.76 | Improved |
| increase channel threshold from 0.5 to 0.7 (ablation) | DPR-Net · S3DIS 2-way 1-shot S0 | mIoU: 80.76 -> 77.82 | 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.
- DyPolySeg — DyPolySeg: Taylor Series-Inspired Dynamic Polynomial Fitting Network for Few-shot Point Cloud Semantic Segmentation (ICML 2025) · better on 7 of 8 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
- Generated and Pseudo Content guided Prototype Refinement for Few-shot Point Cloud Segmentation (NeurIPS 2024 · similarity 0.92)
- Reasoning Beyond Points: A Visual Introspective Approach for Few-Shot 3D Segmentation (NeurIPS 2025 · similarity 0.89)
- DyPolySeg: Taylor Series-Inspired Dynamic Polynomial Fitting Network for Few-shot Point Cloud Semantic Segmentation (ICML 2025 · similarity 0.89)
- Rethinking Few-shot 3D Point Cloud Semantic Segmentation (CVPR 2024 · similarity 0.88)
- White Aggregation and Restoration for Few-shot 3D Point Cloud Semantic Segmentation (ECCV 2026 · similarity 0.88)
- Taylor Series-Inspired Local Structure Fitting Network for Few-shot Point Cloud Semantic Segmentation (AAAI 2025 · 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.
