Research map / Point clouds / Segmentation and detection
SemanticKITTI: A Dataset for Semantic Scene Understanding of LiDAR Sequences
Jens Behley, Martin Garbade, Andres Milioto, Jan Quenzel, Sven Behnke, Cyrill Stachniss, Jurgen Gall
Read the paperCheck your idea against 137,217 papers
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 |
|---|---|---|---|
| replace SqueezeSeg backbone with DarkNet21 (25M parameters) | SqueezeSeg · SemanticKITTI single scan (19 classes, test seq 11-21) | mIoU: 29.5 -> 47.4 | Improved |
| deepen backbone from DarkNet21 to DarkNet53 (50M parameters) | DarkNet21Seg · SemanticKITTI single scan (19 classes, test seq 11-21) | mIoU: 47.4 -> 49.9 | Improved |
| use 5 aggregated scans instead of a single scan | DarkNet53Seg · SemanticKITTI (25 classes in multi-scan) | mIoU: 49.9 -> 41.6 | Mixed |
| add RGB-based 2D semantic segmentation (TS3D) to SSCNet | SSCNet · SemanticKITTI semantic scene completion (test) | semantic scene completion mIoU: 9.53 -> 9.54 | No clear effect |
| use DarkNet53Seg point labels in TS3D | TS3D · SemanticKITTI semantic scene completion (test) | semantic scene completion mIoU: 9.54 -> 10.19 | Improved |
| use DarkNet53Seg point labels in TS3D | TS3D · SemanticKITTI semantic scene completion (test) | scene completion IoU: 29.81 -> 24.99 | Hurt |
| replace SSCNet 3D-CNN backbone with SATNet | TS3D + DarkNet53Seg · SemanticKITTI semantic scene completion (test) | semantic scene completion mIoU: 10.19 -> 17.70 | Improved |
| replace SSCNet 3D-CNN backbone with SATNet | TS3D + DarkNet53Seg · SemanticKITTI semantic scene completion (test) | scene completion IoU: 24.99 -> 50.60 | Improved |
| increase distance to sensor | all single-scan methods · SemanticKITTI single scan (19 classes, test seq 11-21) | mIoU | Hurt |
Closest papers by meaning
- ALSO: Automotive Lidar Self-Supervision by Occupancy Estimation (CVPR 2023 · similarity 0.92)
- 4D Panoptic LiDAR Segmentation (CVPR 2021 · similarity 0.91)
- Temporal Consistent 3D LiDAR Representation Learning for Semantic Perception in Autonomous Driving (CVPR 2023 · similarity 0.91)
- TALoS: Enhancing Semantic Scene Completion via Test-time Adaptation on the Line of Sight (NeurIPS 2024 · similarity 0.90)
- HiLoTs: High-Low Temporal Sensitive Representation Learning for Semi-Supervised LiDAR Segmentation in Autonomous Driving (CVPR 2025 · similarity 0.89)
- Exploring Scene Affinity for Semi-Supervised LiDAR Semantic Segmentation (CVPR 2025 · similarity 0.89)
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.
