Research map / Point clouds / Segmentation and detection
Point-to-Voxel Knowledge Distillation for LiDAR Semantic Segmentation
Check your idea against 108,967 papers
AI-extracted summary. Written by AI from the paper's title only (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 |
|---|---|---|---|
| Apply PVD distillation to the pruned Cylinder3D student | Cylinder3D_0.5× · SemanticKITTI test | mIoU: 65.3 -> 68.9 | Helped |
| Apply PVD distillation to the pruned Cylinder3D student | Cylinder3D_0.5× · nuScenes validation | mIoU: 73.6 -> 76.0 | Helped |
| Apply PVD distillation to pruned SPVNAS | SPVNAS_0.5× · SemanticKITTI validation | mIoU: 60.4 -> 63.8 | Helped |
| Apply PVD distillation to pruned MinkowskiNet | MinkowskiNet_0.5× · SemanticKITTI validation | mIoU: 58.9 -> 61.8 | Helped |
| Add voxelwise output distillation to pointwise output distillation (ablation) | PVD on Cylinder3D_0.5× · SemanticKITTI validation | mIoU: 63.4 -> 64.1 | Helped |
| Add inter-point affinity distillation to both output losses (ablation) | PVD on Cylinder3D_0.5× · SemanticKITTI validation | mIoU: 64.1 -> 64.7 | Helped |
| Add inter-voxel affinity distillation (ablation) | PVD on Cylinder3D_0.5× · SemanticKITTI validation | mIoU: 64.7 -> 66.4 | Helped |
| Set supervoxel size to (120, 60, 8) rather than smaller or larger tested sizes (ablation) | PVD on Cylinder3D_0.5× · SemanticKITTI validation | mIoU: 65.3, 65.7, 66.4, 65.6, 65.2 | Helped |
| Use difficulty-aware sampling rather than distance-aware, category-aware, or random sampling (ablation) | PVD affinity distillation · SemanticKITTI validation | mIoU | Helped |
| Use the standard coefficients (0.1, 0.15, 0.15, 0.25) rather than equal coefficients (ablation) | PVD on Cylinder3D_0.5× · SemanticKITTI validation | mIoU: 65.2 -> 66.4 | Helped |
| Swap point-based and voxel-based loss coefficients (ablation) | PVD on Cylinder3D_0.5× · SemanticKITTI validation | mIoU: 66.4 -> 65.4 | Hurt |
| Vary the class-frequency sampling parameters across the tested choices (ablation) | PVD on Cylinder3D_0.5× · SemanticKITTI validation | mIoU: 66.2 to 66.4 | No clear effect |
Closest papers by meaning
- LaserMix for Semi-Supervised LiDAR Semantic Segmentation (CVPR 2023 · similarity 0.91)
- Annotator: A Generic Active Learning Baseline for LiDAR Semantic Segmentation (NeurIPS 2023 · similarity 0.91)
- Exploring Scene Affinity for Semi-Supervised LiDAR Semantic Segmentation (CVPR 2025 · similarity 0.89)
- Fine-grained Image-to-LiDAR Contrastive Distillation with Visual Foundation Models (NeurIPS 2024 · similarity 0.88)
- Three Pillars Improving Vision Foundation Model Distillation for Lidar (CVPR 2024 · similarity 0.88)
- Construct to Associate: Cooperative Context Learning for Domain Adaptive Point Cloud Segmentation (CVPR 2024 · similarity 0.88)
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.