Research map / Probabilistic and scientific modeling / Uncertainty and Gaussian processes
Analytic Bayesian Uncertainty for LiDAR Segmentation: A Single-pass Generative Approach
Hanieh Shojaei Miandashti, Qianqian Zou, Claus Brenner
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 NIG priors over GMM parameters (GMM-NIG vs GMMSeg) | GMMSeg with RangeFormer · SemanticKITTI val | ACE (%): 3.88 -> 2.01 | Improved |
| add NIG priors over GMM parameters (GMM-NIG vs GMMSeg) | GMMSeg with RangeViT · SemanticKITTI val | ACE (%): 6.02 -> 2.68 | Improved |
| add NIG priors over GMM parameters (GMM-NIG vs GMMSeg) | GMMSeg with SalsaNext · SemanticKITTI val | mIoU (%): 57.01 -> 57.96 | Improved |
| replace softmax confidence with GMM-NIG posterior predictive | RangeFormer · SemanticKITTI val | mIoU (%): 63.60 -> 65.71 | Improved |
| use GMM-NIG instead of deep ensembles for failure detection | RangeFormer · SemanticKITTI val | AUROC (%): 87.61 -> 89.01 | Improved |
| use GMM-NIG instead of deep ensembles for failure detection | RangeFormer · SemanticKITTI val | AUPR (%): 48.07 -> 53.33 | Improved |
| add post-hoc Dirichlet scaling to MSP | RangeViT · SemanticKITTI val | ACE (%): 5.51 -> 6.91 | Hurt |
| add EDL to MSP baseline | RangeFormer · SemanticKITTI val | mIoU (%): 63.60 -> 63.51 | No clear effect |
| attach GMM-NIG head to PTv3 | PTv3 · SemanticKITTI val | ACE (%): 6.81 -> 3.09 | Improved |
| attach GMM-NIG head to SphereFormer | SphereFormer · SemanticKITTI val | ACE (%): 4.67 -> 2.08 | Improved |
| use low-rank covariance (R=8) instead of factorized (ablation) | GMM-NIG SalsaNext · SemanticKITTI val | mIoU (%): 57.96 -> 58.01 | No clear effect |
| use low-rank covariance (R=8) instead of factorized (ablation) | GMM-NIG SalsaNext · SemanticKITTI val | ACE (%): 2.10 -> 2.87 | Hurt |
Closest papers by meaning
- Toward Enhanced Representation Learning for Single-Source Domain Generalization in LiDAR Semantic Segmentation (TPAMI 2026 · similarity 0.88)
- PC-BEV: An Efficient Polar-Cartesian BEV Fusion Framework for LiDAR Semantic Segmentation (AAAI 2025 · similarity 0.85)
- OLiDM: Object-aware LiDAR Diffusion Models for Autonomous Driving (AAAI 2025 · similarity 0.84)
- UniMix: Towards Domain Adaptive and Generalizable LiDAR Semantic Segmentation in Adverse Weather (CVPR 2024 · similarity 0.84)
- Predictive Uncertainty Quantification for Bird's Eye View Segmentation: A Benchmark and Novel Loss Function (ICLR 2025 · similarity 0.84)
- Multi-Space Alignments Towards Universal LiDAR Segmentation (CVPR 2024 · 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.
