Research map / Detection and segmentation / 3D detection and driving
ALSO: Automotive Lidar Self-Supervision by Occupancy Estimation
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 |
|---|---|---|---|
| Pre-train with visibility-derived local occupancy reconstruction | MinkUNet · nuScenes 1% labeled | mIoU (%): 35.0 -> 37.4 | Helped |
| Pre-train with visibility-derived local occupancy reconstruction | MinkUNet · SemanticKITTI 0.1% labeled | mIoU (%): 32.5 -> 35.0 | Helped |
| Pre-train with visibility-derived local occupancy reconstruction | MinkUNet · LivoxSimu 1% labeled | mIoU (%): 63.8 -> 65.5 | Helped |
| Initialize the detector with ALSO pre-training on KITTI-360 | SECOND on KITTI3D · KITTI3D validation, moderate difficulty, R40 | mAP (%): 65.35 -> 68.31 | Helped |
| Initialize the detector with ALSO pre-training on KITTI-360 | PV-RCNN on KITTI3D · KITTI3D validation, moderate difficulty, R40 | mAP (%): 70.57 -> 72.96 | Helped |
| Add intensity as an input feature during pre-training (ablation) | ALSO · nuScenes ablation validation, 1% labeled | mIoU (%): 36.4 -> 38.2 | Helped |
| Add intensity reconstruction loss (ablation) | ALSO with intensity input · nuScenes ablation validation, 1% labeled | mIoU (%): 38.2 -> 38.4 | Helped |
| Use a larger reconstruction radius of 4 m rather than 1 m (ablation) | ALSO · nuScenes ablation validation, 1% labeled | mIoU (%): 38.4 -> 36.4 | Hurt |
| Use a reconstruction radius of 0.5 m rather than 1 m (ablation) | ALSO · nuScenes ablation validation, 1% labeled | mIoU (%): 38.4 -> 37.6 | Hurt |
| Set the behind-surface offset delta to 0.2 m or 0.4 m rather than 0.1 m (ablation) | ALSO · nuScenes ablation validation, 1% labeled | mIoU (%): 38.4 -> 38.7 | Helped |
| Replace the per-point occupancy decoder with a POCO KNN plus attention decoder (ablation) | ALSO · nuScenes ablation validation, 1% labeled | mIoU (%): 38.4 -> 33.8 | Hurt |
| Replace the per-point decoder with ball search and average pooling (ablation) | ALSO · nuScenes ablation validation, 1% labeled | mIoU (%): 38.4 -> 35.7 | Hurt |
Closest papers by meaning
- VLM2Scene: Self-Supervised Image-Text-LiDAR Learning with Foundation Models for Autonomous Driving Scene Understanding (AAAI 2024 · similarity 0.90)
- UnO: Unsupervised Occupancy Fields for Perception and Forecasting (CVPR 2024 · similarity 0.89)
- HiLoTs: High-Low Temporal Sensitive Representation Learning for Semi-Supervised LiDAR Segmentation in Autonomous Driving (CVPR 2025 · similarity 0.88)
- Exploring Scene Affinity for Semi-Supervised LiDAR Semantic Segmentation (CVPR 2025 · similarity 0.88)
- Is Your LiDAR Placement Optimized for 3D Scene Understanding? (NeurIPS 2024 · similarity 0.88)
- OLiDM: Object-aware LiDAR Diffusion Models for Autonomous Driving (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.