Research map / Label-efficient and robust learning / Test-time adaptation
Adapting Point Cloud Analysis via Multimodal Bayesian Distribution Learning
Xingyu Zhu, Liang Yi, Shuo Wang, Wenbo Zhu, Yongliang Wu, Beier Zhu, Hanwang Zhang
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 geometric distribution learning to fixed text prototypes (ablation) | BayesMM on ULIP-2 · S-OBJ | accuracy: 42.00 -> 46.47 | Improved |
| add geometric distribution learning to fixed text prototypes (ablation) | BayesMM on ULIP-2 · Omni3D | accuracy: 26.58 -> 26.63 | No clear effect |
| add textual distribution learning instead (ablation) | BayesMM on ULIP-2 · S-OBJ | accuracy: 42.00 -> 52.50 | Improved |
| add textual distribution learning instead (ablation) | BayesMM on ULIP-2 · Omni3D | accuracy: 26.58 -> 30.79 | Improved |
| add geometric distribution and Bayesian weighting on top of textual distribution (ablation) | BayesMM on ULIP-2 · S-OBJ | accuracy: 52.50 -> 53.02 | Improved |
| add geometric distribution and Bayesian weighting on top of textual distribution (ablation) | BayesMM on ULIP-2 · Omni3D | accuracy: 30.79 -> 31.12 | Improved |
| vary alpha^2 weighting strength | BayesMM · ModelNet-C | accuracy: minimal impact | No clear effect |
| vary beta^2 variance scale | BayesMM · ModelNet-C | accuracy: ULIP stays around 59.4%, ULIP-2 near 65.2% | No clear effect |
| apply BayesMM instead of Point-Cache (Hierarchical) | ULIP · ModelNet-C avg | accuracy: 55.02 -> 59.42 | Improved |
| apply BayesMM to zero-shot backbone | ULIP · ModelNet-C avg | accuracy: 48.60% to 59.42% (+10.82) | Improved |
| apply BayesMM to zero-shot backbone | Uni3D · Generalization avg | accuracy: 54.09 -> 57.72 | Improved |
| apply BayesMM instead of Point-Cache (Hierarchical) | Uni3D · Sim-to-real avg | accuracy: 69.07 -> 70.94 | Mixed |
Closest papers by meaning
- Point-Cache: Test-time Dynamic and Hierarchical Cache for Robust and Generalizable Point Cloud Analysis (CVPR 2025 · similarity 0.88)
- PointMAC: Meta-Learned Adaptation for Robust Test-Time Point Cloud Completion (NeurIPS 2025 · similarity 0.86)
- SMART-PC: Skeletal Model Adaptation for Robust Test-Time Training in Point Clouds (ICML 2025 · similarity 0.86)
- Adaptive 3D Reconstruction via Diffusion Priors and Forward Curvature-Matching Likelihood Updates (NeurIPS 2025 · similarity 0.86)
- Towards Robust Multimodal Open-set Test-time Adaptation via Adaptive Entropy-aware Optimization (ICLR 2025 · similarity 0.85)
- PCoTTA: Continual Test-Time Adaptation for Multi-Task Point Cloud Understanding (NeurIPS 2024 · similarity 0.85)
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.
