Research map / Continual learning / Class-incremental
Point-UQ: An Uncertainty-Quantification Paradigm for Point Cloud Few-Shot Class Incremental Learning
Xiangqi Li, Libo Huang, Jiarui Zhao, Weilun Feng, Chuanguang Yang, Zhulin An, Yongjun Xu
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 AAE multi-scale attention fusion (ablation) | Point-UQ (with UDD) · ShapeNet to CO3D cross-dataset | harmonic accuracy average: 60.0 -> 64.8 | Improved |
| add UDD entropy-based decision fusion (ablation) | Point-UQ (with AAE) · ShapeNet to CO3D cross-dataset | harmonic accuracy average: 55.7 -> 64.8 | Improved |
| add both AAE and UDD (ablation) | baseline without modules · ShapeNet to CO3D cross-dataset | harmonic accuracy average: 54.5 -> 64.8 | Improved |
| replace fixed-rule fusion with self-attention fusion (ablation) | Deep-Semantic-only · ShapeNet to CO3D cross-dataset | harmonic accuracy average: 60.0 -> 64.8 | Improved |
| replace LayerWise-To-Last fusion with self-attention fusion (ablation) | Point-UQ · ShapeNet to CO3D cross-dataset | harmonic accuracy average: 61.0 -> 64.8 | Improved |
| replace Symmetric-Cross-Fusion with self-attention fusion (ablation) | Point-UQ · ShapeNet to CO3D cross-dataset | harmonic accuracy average: 59.4 -> 64.8 | Improved |
| use sigmoid instead of softplus_norm activation (ablation) | UDD weighting · ShapeNet to CO3D cross-dataset | relative accuracy drop: 18.1 -> 12.1 | Improved |
| use sigmoid instead of relu_sigmoid activation (ablation) | UDD weighting · ShapeNet to CO3D cross-dataset | relative accuracy drop: 12.8 -> 12.1 | No clear effect |
| use entropy instead of semantic confidence gap as uncertainty (ablation) | UDD · ShapeNet to CO3D cross-dataset | relative accuracy drop: 13.7 -> 12.1 | Improved |
| use entropy instead of semantic-geometric consistency as uncertainty (ablation) | UDD · ShapeNet to CO3D cross-dataset | relative accuracy drop: 13.1 -> 12.1 | Improved |
| raise lambda from 0.5 to 2.0 (ablation) | UDD entropy mapping · ShapeNet to CO3D cross-dataset | harmonic accuracy average: 10.9 -> 64.8 | Improved |
| raise lambda from 2.0 to 3.0 (ablation) | UDD entropy mapping · ShapeNet to CO3D cross-dataset | harmonic accuracy average: 64.8 -> 66.5 | Mixed |
Closest papers by meaning
- Hyperbolic Uncertainty-Aware Few-Shot Incremental Point Cloud Segmentation (CVPR 2025 · similarity 0.89)
- Geometric Feature Embedding for Effective 3D Few-Shot Class Incremental Learning (ICML 2025 · similarity 0.88)
- Generalized Zero-Shot Learning for Point Cloud Segmentation with Evidence-Based Dynamic Calibration (AAAI 2025 · similarity 0.87)
- Activating Sparse Part Concepts for 3D Class Incremental Learning (CVPR 2025 · similarity 0.86)
- Adaptive Decision Boundary for Few-Shot Class-Incremental Learning (AAAI 2025 · similarity 0.86)
- Tripartite Weight-Space Ensemble for Few-Shot Class-Incremental Learning (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.
