Research map / Point clouds / Segmentation and detection
LiDAR Snowfall Simulation for Robust 3D Object Detection
Martin Hahner, Christos Sakaridis, Mario Bijelic, Felix Heide, Fisher Yu, Dengxin Dai, Luc Van Gool
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 |
|---|---|---|---|
| Train PV-RCNN with snowfall and wet-ground simulation instead of no weather augmentation | PV-RCNN · STF heavy snowfall, 0-80m | 3D AP: 39.69 -> 41.79 | Improved |
| Train all seven detectors with snowfall and wet-ground simulation instead of clear-weather data only | PV-RCNN, VoxelRCNN-Car, CenterPoint, Part-A², PointRCNN, SECOND, PointPillars · STF heavy snowfall, 0-80m | 3D AP | Improved |
| Add wet-ground simulation to snowfall simulation (ablation) | PV-RCNN · STF heavy snowfall, 0-80m | 3D AP: 41.61 -> 41.79 | No clear effect |
| Add wet-ground simulation to snowfall simulation (ablation) | VoxelRCNN-Car · STF heavy snowfall, 0-80m | 3D AP: 41.20 -> 40.76 | Hurt |
| Add wet-ground simulation to snowfall simulation (ablation) | CenterPoint · STF heavy snowfall, 0-80m | 3D AP: 39.81 -> 40.14 | Improved |
| Add wet-ground simulation to snowfall simulation (ablation) | Part-A² · STF heavy snowfall, 0-80m | 3D AP: 37.73 -> 37.52 | Hurt |
| Add wet-ground simulation to snowfall simulation (ablation) | PointRCNN · STF heavy snowfall, 0-80m | 3D AP: 37.59 -> 37.51 | No clear effect |
| Add wet-ground simulation to snowfall simulation (ablation) | SECOND · STF heavy snowfall, 0-80m | 3D AP: 36.83 -> 37.03 | No clear effect |
| Add wet-ground simulation to snowfall simulation (ablation) | PointPillars · STF heavy snowfall, 0-80m | 3D AP: 32.94 -> 31.38 | Hurt |
| Train with the proposed snowfall simulation without wet-ground simulation instead of no augmentation (ablation) | PV-RCNN · STF clear weather, 0-80m | 3D AP: 45.36 -> 45.61 | Improved |
| Use the proposed snowfall simulation rather than the clear-weather baseline (ablation) | six of seven detection methods · STF clear weather, 0-80m | 3D AP | Improved |
| Apply fog simulation as training augmentation rather than no augmentation | most detection methods · STF heavy and light snowfall | 3D AP | Hurt |
Closest papers by meaning
- Sunshine to Rainstorm: Cross-Weather Knowledge Distillation for Robust 3D Object Detection (AAAI 2024 · similarity 0.88)
- An Empirical Study of the Generalization Ability of Lidar 3D Object Detectors to Unseen Domains (CVPR 2024 · similarity 0.86)
- Equirectangular Point Reconstruction for Domain Adaptive Multimodal 3D Object Detection in Adverse Weather Conditions (AAAI 2025 · similarity 0.86)
- Towards Robust 3D Object Detection with LiDAR and 4D Radar Fusion in Various Weather Conditions (CVPR 2024 · similarity 0.86)
- L4DR: LiDAR-4DRadar Fusion for Weather-Robust 3D Object Detection (AAAI 2025 · similarity 0.85)
- UAWTrack: Universal 3D Single Object Tracking in Adverse Weather (AAAI 2025 · 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.
