Research map / Point clouds / Segmentation and detection
Self-Prediction for Joint Instance and Semantic Segmentation of Point Clouds
Jinxian Liu, Minghui Yu, Bingbing Ni⁴, Ye Chen
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 instance Self-Prediction head (ablation) | PointNet++ baseline · S3DIS Area 5 (instance seg) | mPrec: 53.4 -> 58.9 | Improved |
| add semantic Self-Prediction head (ablation) | PointNet++ baseline · S3DIS Area 5 (semantic seg) | mIoU: 50.8 -> 55.9 | Improved |
| add semantic Self-Prediction on top of instance Self-Prediction (joint multi-label) (ablation) | PointNet++ Ins-SP · S3DIS Area 5 (instance seg) | mPrec: 58.9 -> 60.1 | Improved |
| add instance Self-Prediction on top of semantic Self-Prediction (joint multi-label) (ablation) | PointNet++ Sem-SP · S3DIS Area 5 (semantic seg) | mIoU: 55.9 -> 58.8 | Improved |
| use joint Self-Prediction instead of baseline | PointNet · S3DIS Area 5 (semantic seg) | mIoU: 44.7 -> 52.4 | Improved |
| replace ASIS association module with Self-Prediction | PointNet++ · S3DIS Area 5 (instance seg) | mPrec: 55.3 -> 60.1 | Improved |
| use bidirectional instead of unidirectional propagation (ablation) | InsSem-SP, PointNet · S3DIS Area 5 | mPrec: 49.9 -> 51.1 | Improved |
| use bidirectional instead of unidirectional propagation (ablation) | InsSem-SP, PointNet · S3DIS Area 5 | mIoU: 51.0 -> 52.4 | Improved |
| divide groups by instance labels instead of randomly (ablation) | InsSem-SP, PointNet · S3DIS Area 5 | mPrec: 50.5 -> 51.1 | Improved |
| divide groups by instance labels instead of randomly (ablation) | InsSem-SP, PointNet · S3DIS Area 5 | mIoU: 51.6 -> 52.4 | Improved |
| add Self-Prediction to DGCNN backbone | DGCNN · ShapeNet part | pIoU: 85.2 -> 86.2 | Improved |
| add Self-Prediction to PointNet++ backbone | PointNet++ · ShapeNet part | pIoU: 84.5 -> 85.8 | Improved |
Closest papers by meaning
- One Thing One Click: A Self-Training Approach for Weakly Supervised 3D Semantic Segmentation (CVPR 2021 · similarity 0.91)
- Associatively Segmenting Instances and Semantics in Point Clouds (CVPR 2019 · similarity 0.91)
- Dual-level Adaptive Self-Labeling for Novel Class Discovery in Point Cloud Segmentation (ECCV 2024 · similarity 0.90)
- Relation3D : Enhancing Relation Modeling for Point Cloud Instance Segmentation (CVPR 2025 · similarity 0.89)
- FreePoint: Unsupervised Point Cloud Instance Segmentation (CVPR 2024 · similarity 0.88)
- OneFormer3D: One Transformer for Unified Point Cloud Segmentation (CVPR 2024 · 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.
