Research map / Point clouds / Segmentation and detection
Serialization based Point Cloud Oversegmentation
Chenghui Lu, Jianlong Kwan, Dilong Li, Ziyi Chen, Haiyan Guan
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 |
|---|---|---|---|
| add Mamba at the U-Net bottleneck (ablation) | 3D sparse U-Net · ScanNet val | BR (%): 76.33 -> 84.23 | Improved |
| add Mamba at the U-Net bottleneck (ablation) | 3D sparse U-Net · ScanNet val | mIoU (%): 74.10 -> 76.96 | Improved |
| use Mamba only at bottleneck instead of full network (ablation) | SPCNet backbone · ScanNet val | BR (%): 78.37 -> 84.23 | Improved |
| use neighbor expansion of 2 instead of none (ablation) | SPCNet · ScanNet val | BR (%): 79.86 -> 84.23 | Improved |
| increase neighbor expansion from 2 to 6 (ablation) | SPCNet · ScanNet val | BR (%): 84.23 -> 79.93 | Hurt |
| increase neighbor expansion from 2 to 4 (ablation) | SPCNet · ScanNet val | BR (%): 84.23 -> 83.67 | Hurt |
| add level-1 superpoint features (ablation) | backbone for semantic segmentation · ScanNet val | mIoU (%): 76.96 -> 77.50 | Improved |
| add level-2 superpoint features (ablation) | backbone + level-1 · ScanNet val | mIoU (%): 77.50 -> 77.80 | Improved |
| replace SuperLiDAR with SPCNet superpoints | oversegmentation · SemanticKITTI | BR (%): 65.52 -> 73.18 | Improved |
| replace SPNet with SPCNet | oversegmentation · S3DIS Area 5 | BP (%): 13.14 -> 26.78 | Improved |
| replace PTv3 with SPCNet | semantic segmentation · nuScenes val | mIoU (%): 80.4 -> 77.6 | Hurt |
| replace SuperLiDAR with SPCNet | inference speed · SemanticKITTI | time (ms): 72 -> 12 | Improved |
Closest papers by meaning
- Point Cloud Oversegmentation With Graph-Structured Deep Metric Learning (CVPR 2019 · similarity 0.92)
- Superpoint Network for Point Cloud Oversegmentation (ICCV 2021 · similarity 0.91)
- Efficient LiDAR Point Cloud Oversegmentation Network (ICCV 2023 · similarity 0.89)
- Clustering based Point Cloud Representation Learning for 3D Analysis (ICCV 2023 · similarity 0.87)
- Geometrically-driven Aggregation for Zero-shot 3D Point Cloud Understanding (CVPR 2024 · similarity 0.85)
- FreePoint: Unsupervised Point Cloud Instance Segmentation (CVPR 2024 · 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.
