Research map / Point clouds / Segmentation and detection
Point Cloud Instance Segmentation With Semi-Supervised Bounding-Box Mining
Yongbin Liao, Hongyuan Zhu, Yanggang Zhang, Chuangguan Ye, Tao Chen, Jiayuan Fan
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 semi-supervised SPCR bounding box proposals instead of VoteNet trained on labeled data only (ablation) | VoteNet · ScanNet v2 val (10% labeled) | mAP@25%: 26.3 -> 31.4 | Improved |
| add SPCR on data-efficient detection benchmark versus CSC (ablation) | SPCR · ScanNet data-efficient detection (10% data) | mAP@50%: 8.6 -> 18.0 | Improved |
| add SPCR at 80% data versus CSC (ablation) | SPCR · ScanNet data-efficient detection (80% data) | mAP@50%: 36.7 -> 36.4 | No clear effect |
| add SPIB (semantic propagation, property graph, occupancy refinement) to VoteNet at 10% labels | VoteNet · ScanNet v2 val (10% labeled) | mAP@25%: 26.3 -> 38.8 | Improved |
| add semantic propagation to DetIns baseline (ablation) | DetIns · ScanNet v2 val | mAP@25%: 56.9 -> 57.2 | Improved |
| add property consistency graph to semantic propagation (ablation) | DetIns + SP · ScanNet v2 val | mAP@25%: 57.2 -> 59.1 | Improved |
| add occupancy ratio guided refinement (ablation) | DetIns + SP + PC · ScanNet v2 val | mAP@25%: 59.1 -> 61.4 | Improved |
| combine coordinate, color and normal property terms instead of coordinate alone (ablation) | property consistency graph · ScanNet v2 val | mAP@25%: 55.1 -> 57.5 | Improved |
| combine coordinate, color and normal property terms instead of coordinate alone (ablation) | property consistency graph · ScanNet v2 val | mAP@50%: 32.6 -> 34.1 | Improved |
| add color to coordinate property term (ablation) | property consistency graph · ScanNet v2 val | mAP@25%: 55.1 -> 56.1 | Improved |
| use box-level supervision with SPIB instead of DSRG baseline (ablation) | DSRG · ScanNet v2 val | mAP@25%: 35.1 -> 61.4 | Improved |
| use SPIB instead of BoxMask baseline (ablation) | BoxMask · ScanNet v2 val | mAP@25%: 58.4 -> 61.4 | Improved |
Closest papers by meaning
- GaPro: Box-Supervised 3D Point Cloud Instance Segmentation Using Gaussian Processes as Pseudo Labelers (ICCV 2023 · similarity 0.89)
- Sketchy Bounding-box Supervision for 3D Instance Segmentation (CVPR 2025 · similarity 0.89)
- FreePoint: Unsupervised Point Cloud Instance Segmentation (CVPR 2024 · similarity 0.88)
- Box2Mask: Weakly Supervised 3D Semantic Instance Segmentation Using Bounding Boxes (ECCV 2022 · similarity 0.88)
- Learning Object Bounding Boxes for 3D Instance Segmentation on Point Clouds (NeurIPS 2019 · similarity 0.88)
- Extreme Point Supervised Instance Segmentation (CVPR 2024 · similarity 0.86)
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.
