Research map / Point clouds / Segmentation and detection
3D Semantic Segmentation in the Wild: Learning Generalized Models for Adverse-Condition Point Clouds
Aoran Xiao, Jiaxing Huang, Weihao Xuan, Ruijie Ren, Kangcheng Liu, Dayan Guan, Abdulmotaleb El Saddik, Shijian Lu, Eric P. Xing
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 |
|---|---|---|---|
| Use PointDR's geometry randomization and embedding aggregation instead of the source-only baseline | MinkowskiNet · SemanticKITTI→SemanticSTF | mIoU: 24.4 -> 28.6 | Improved |
| Use PointDR instead of the strongest listed augmentation baseline | MinkowskiNet · SynLiDAR→SemanticSTF | mIoU: 15.7 -> 18.5 | Improved |
| Add contrastive loss to cross-entropy training (ablation) | PointDR · SemanticKITTI→SemanticSTF | mIoU: 24.4 -> 27.4 | Improved |
| Add a momentum-updated class memory bank to contrastive training (ablation) | PointDR-CT · SemanticKITTI→SemanticSTF | mIoU: 27.4 -> 28.6 | Improved |
| Use dropout augmentation (ablation) | Source-only MinkowskiNet · SemanticKITTI→SemanticSTF | mIoU: 24.4 -> 25.7 | Improved |
| Use noise perturbation augmentation (ablation) | Source-only MinkowskiNet · SemanticKITTI→SemanticSTF | mIoU: 24.4 -> 25.9 | Improved |
| Use flipping augmentation (ablation) | Source-only MinkowskiNet · SemanticKITTI→SemanticSTF | mIoU: 24.4 -> 25.2 | Improved |
| Use jittering augmentation (ablation) | Source-only MinkowskiNet · SemanticKITTI→SemanticSTF | mIoU: 24.4 -> 26.9 | Improved |
| Combine dropout, noise perturbation, flipping, and jittering (ablation) | Source-only MinkowskiNet · SemanticKITTI→SemanticSTF | mIoU: 26.9 -> 26.1 | Hurt |
| Set contrastive loss weight λcl to 0.10 rather than 0.0 (ablation) | PointDR · SemanticKITTI→SemanticSTF | mIoU: 24.4 -> 28.6 | Improved |
| Set contrastive loss weight λcl to 1.0 rather than 0.10 (ablation) | PointDR · SemanticKITTI→SemanticSTF | mIoU: 28.6 -> 25.1 | Hurt |
| Set memory-bank momentum m to 0.98 rather than 0.99 (ablation) | PointDR · SemanticKITTI→SemanticSTF | mIoU: 28.6 -> 28.1 | 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.
- PolarMix — PolarMix: A General Data Augmentation Technique for LiDAR Point Clouds (NeurIPS 2022) · better on 1 of 1 shared result
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
- BeyondSparse: Facilitating Mamba to Enhance Cross-Domain 3D Semantic Segmentation in Adverse Weather (AAAI 2026 · similarity 0.91)
- UniMix: Towards Domain Adaptive and Generalizable LiDAR Semantic Segmentation in Adverse Weather (CVPR 2024 · similarity 0.89)
- No Thing, Nothing: Highlighting Safety-Critical Classes for Robust LiDAR Semantic Segmentation in Adverse Weather (CVPR 2025 · similarity 0.89)
- Semantic Segmentation in Multiple Adverse Weather Conditions with Domain Knowledge Retention (AAAI 2024 · similarity 0.88)
- Hyperbolic Uncertainty-Aware Few-Shot Incremental Point Cloud Segmentation (CVPR 2025 · similarity 0.87)
- ERF: A Benchmark Dataset for Robust Semantic Segmentation Under Extreme Rainfall Conditions (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.
