Research map / Point clouds / Segmentation and detection
LaserMix for Semi-Supervised LiDAR Semantic Segmentation
Read the paperCheck 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 LaserMix semi-supervised training | FIDNet range-view segmentation · nuScenes 1% labels | mIoU (%): 49.5 vs 43.8 | Helped |
| Apply LaserMix semi-supervised training | Cylinder3D voxel segmentation · nuScenes 10% labels | mIoU (%): 69.9 vs 66.5 | Helped |
| Add the mixing loss L_mix (ablation) | MeanTeacher baseline on nuScenes · nuScenes val, 1% labels, range view | mIoU (%): +3.5 | Helped |
| Add the mixing loss L_mix (ablation) | MeanTeacher baseline on nuScenes · nuScenes val, 10% labels, range view | mIoU (%): +3.9 | Helped |
| Add mean-teacher consistency loss L_mt to mixing (ablation) | Student-supervised LaserMix · nuScenes val, 10% labels, range view | mIoU (%): 64.3 -> 65.5 | Helped |
| Use Teacher instead of Student net supervision for mixed pseudo-labels (ablation) | LaserMix · nuScenes val, 10% labels, range view | mIoU (%): 65.5 -> 68.2 | Helped |
| Use random-point partitioning (MixUp-like) instead of laser partitions (ablation) | LaserMix framework · nuScenes val, 10% labels, range view | mIoU (%) | No clear effect |
| Use laser-beam partitions instead of random-area CutMix partitions (ablation) | LiDAR SSL mixing · nuScenes, SemanticKITTI, and ScribbleKITTI | mIoU (%): up to 3.3% mIoU | Helped |
| Replace dataset-sampled outside-area filling with dummy filling (CutOut) (ablation) | LaserMix mixing · LiDAR SSL benchmark | mIoU (%) | Hurt |
| Reverse laser-area ordering (ablation) | LaserMix · nuScenes val, 10% labels, range view | mIoU (%): 68.2 -> 64.4 (-3.8% mIoU) | Hurt |
| Shuffle laser-area ordering (ablation) | LaserMix · nuScenes val, 10% labels, range view | mIoU (%): 68.2 -> 63.8 (-4.4% mIoU) | Hurt |
| Mix two unlabeled scans rather than labeled with unlabeled (ablation) | LaserMix · nuScenes val, 10% labels, range view | mIoU (%): 68.2 -> 66.9 (-1.3% mIoU) | Hurt |
Closest papers by meaning
- Collaborative Learning for Semi-Supervised LiDAR Semantic Segmentation (ICML 2026 · similarity 0.92)
- Scribble-Supervised LiDAR Semantic Segmentation (CVPR 2022 · similarity 0.92)
- Beyond the Label Itself: Latent Labels Enhance Semi-supervised Point Cloud Panoptic Segmentation (AAAI 2024 · similarity 0.90)
- Exploring Scene Affinity for Semi-Supervised LiDAR Semantic Segmentation (CVPR 2025 · similarity 0.90)
- HiLoTs: High-Low Temporal Sensitive Representation Learning for Semi-Supervised LiDAR Segmentation in Autonomous Driving (CVPR 2025 · similarity 0.88)
- PC-BEV: An Efficient Polar-Cartesian BEV Fusion Framework for LiDAR Semantic Segmentation (AAAI 2025 · 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.