Research map / Point clouds / Segmentation and detection
Learning Object Bounding Boxes for 3D Instance Segmentation on Point Clouds
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 |
|---|---|---|---|
| add the box score prediction sub-branch (ablation) | 3D-BoNet · S3DIS Area 5 | mPrec: 50.9 -> 57.5 | Helped |
| use sIoU alone rather than Euclidean distance alone for box association and loss (ablation) | 3D-BoNet · S3DIS Area 5 | mPrec: 53.8 -> 55.2 | Helped |
| use sIoU alone rather than cross-entropy score alone for box association and loss (ablation) | 3D-BoNet · S3DIS Area 5 | mPrec: 51.8 -> 55.2 | Helped |
| combine Euclidean distance, sIoU, and cross-entropy score instead of using sIoU alone (ablation) | 3D-BoNet · S3DIS Area 5 | mPrec: 55.2 -> 57.5 | Helped |
| add box prediction supervision (ablation) | 3D-BoNet · S3DIS Area 5 | mPrec: 37.3 -> 57.5 | Helped |
| use focal loss instead of standard cross-entropy for point mask prediction (ablation) | 3D-BoNet · S3DIS Area 5 | mPrec: 50.8 -> 57.5 | Helped |
| use 3D-BoNet rather than the strongest reported baseline | ScanNet(v2) · ScanNet(v2) | AP@0.5 (mean): 48.1 -> 48.8 | Helped |
| use 3D-BoNet rather than ASIS | S3DIS · S3DIS | mPrec: 63.6 -> 65.6 | Helped |
| use 3D-BoNet rather than ASIS | S3DIS · S3DIS | mRec: 47.5 -> 47.6 | No clear effect |
| use 3D-BoNet rather than ASIS | ScanNet(v2) validation split · ScanNet(v2) validation split | total time: 56757 -> 2871 | Helped |
Closest papers by meaning
- Sketchy Bounding-box Supervision for 3D Instance Segmentation (CVPR 2025 · similarity 0.90)
- Spherical Mask: Coarse-to-Fine 3D Point Cloud Instance Segmentation with Spherical Representation (CVPR 2024 · similarity 0.88)
- OneFormer3D: One Transformer for Unified Point Cloud Segmentation (CVPR 2024 · similarity 0.88)
- FreePoint: Unsupervised Point Cloud Instance Segmentation (CVPR 2024 · similarity 0.87)
- Improved MLP Point Cloud Processing with High-Dimensional Positional Encoding (AAAI 2024 · similarity 0.87)
- Insightful Instance Features for 3D Instance Segmentation (CVPR 2025 · similarity 0.87)
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.