No Thing, Nothing: Highlighting Safety-Critical Classes for Robust LiDAR Semantic Segmentation in Adverse Weather

2025-03-20

AI Summary

Existing LiDAR semantic segmentation methods exhibit significantly lower accuracy for dynamic “things” classes (e.g., vehicles, pedestrians) than for static “stuff” classes under adverse weather conditions, posing a critical safety risk for autonomous driving. To address this safety-critical bottleneck, we propose NTN—a novel method specifically designed to enhance robustness in degraded sensing conditions. First, we introduce a superclass feature binding mechanism that suppresses cross-class misclassification at the semantic level. Second, we develop a laser-beam-local region modeling module coupled with adversarial feature-space alignment regularization to improve local degradation robustness. Evaluated on two challenging cross-domain benchmarks—SemanticKITTI→STF and SemanticPOSS→STF—NTN achieves state-of-the-art performance: overall mIoU improves by 2.6 and 7.9 points, respectively, while things-class mIoU increases substantially by 4.8 and 7.9 points, demonstrating superior generalization under adverse weather.

Technology Category

• Computer Vision: Segmentation
• Natural Language Processing: Safety and Robustness
• Machine Learning: Adversarial Learning & Robustness

Application Category

• Semantics and Knowledge: Methods to enhance, augment, integrate or synergize semantic models such as knowledge graphs and LLMs
• Search and Retrieval-Augmented AI: Vertical and domain-specific search
• Systems and Infrastructure for Web, Mobile and WoT: Applied ML and AI for Web-based mobile applications

Abstract

Existing domain generalization methods for LiDAR semantic segmentation under adverse weather struggle to accurately predict"things"categories compared to"stuff"categories. In typical driving scenes,"things"categories can be dynamic and associated with higher collision risks, making them crucial for safe navigation and planning. Recognizing the importance of"things"categories, we identify their performance drop as a serious bottleneck in existing approaches. We observed that adverse weather induces degradation of semantic-level features and both corruption of local features, leading to a misprediction of"things"as"stuff". To mitigate these corruptions, we suggest our method, NTN - segmeNt Things for No-accident. To address semantic-level feature corruption, we bind each point feature to its superclass, preventing the misprediction of things classes into visually dissimilar categories. Additionally, to enhance robustness against local corruption caused by adverse weather, we define each LiDAR beam as a local region and propose a regularization term that aligns the clean data with its corrupted counterpart in feature space. NTN achieves state-of-the-art performance with a +2.6 mIoU gain on the SemanticKITTI-to-SemanticSTF benchmark and +7.9 mIoU on the SemanticPOSS-to-SemanticSTF benchmark. Notably, NTN achieves a +4.8 and +7.9 mIoU improvement on"things"classes, respectively, highlighting its effectiveness.

Problem

• Improves LiDAR semantic segmentation in adverse weather
• Addresses misprediction of dynamic 'things' as 'stuff'
• Enhances robustness against local and semantic feature corruption

Innovation

• Binds point features to superclasses for accuracy
• Defines LiDAR beams as local regions
• Regularizes clean and corrupted data alignment