Research map / Point clouds / Segmentation and detection
Multi-Path Region Mining for Weakly Supervised 3D Semantic Segmentation on Point Clouds
Check 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 |
|---|---|---|---|
| Use subcloud-level rather than scene-level weak labels | PCAM and MPRM on ScanNet · ScanNet training pseudo labels | mIoU: 22.1 -> 43.1; 24.4 -> 46.6 | Helped |
| Add MPRM paths to the PCAM baseline | ScanNet pseudo labels with scene-level supervision · ScanNet training pseudo labels | mIoU: 22.1 -> 24.4 | Helped |
| Add MPRM paths to the PCAM baseline | ScanNet pseudo labels with subcloud-level supervision · ScanNet training pseudo labels | mIoU: 43.1 -> 46.6 | Helped |
| Use all four paths with max fusion instead of a single path (ablation) | MPRM · ScanNet pseudo labels, training and validation sets | mIoU: 44.3 -> 46.6; 39.3 -> 41.0 | Helped |
| Replace max fusion with sum fusion (ablation) | MPRM with all four paths · ScanNet pseudo labels, training and validation sets | mIoU: 46.6 -> 45.9; 41.0 -> 39.7 | Hurt |
| Add dCRF post-processing to MPRM pseudo labels (ablation) | MPRM with subcloud-level supervision · ScanNet training pseudo labels | mIoU: 46.6 -> 47.4 | Helped |
| Retrain a full-scale segmentation network on dCRF-refined pseudo labels (ablation) | MPRM-CRF · ScanNet validation | mIoU: 41.4 -> 43.2 | Helped |
| Add the spatial attention path to PCAM (ablation) | MPRM · ScanNet pseudo labels, training and validation sets | mIoU: 44.3 -> 46.0; 39.3 -> 40.3 | Helped |
| Add the channel attention path to PCAM (ablation) | MPRM · ScanNet pseudo labels, training and validation sets | mIoU: 44.3 -> 45.9; 39.3 -> 40.0 | Helped |
| Add the point-wise attention path to PCAM (ablation) | MPRM · ScanNet pseudo labels, training and validation sets | mIoU: 44.3 -> 45.6; 39.3 -> 40.4 | Helped |
| Use the spatial attention path instead of the PCAM-only path (ablation) | MPRM · ScanNet pseudo labels, training and validation sets | mIoU: 44.3 -> 44.8; 39.3 -> 39.4 | Mixed |
| Use the channel attention path instead of the PCAM-only path (ablation) | MPRM · ScanNet pseudo labels, training and validation sets | mIoU: 44.3 -> 44.3; 39.3 -> 39.3 | No clear effect |
Closest papers by meaning
- Weakly Supervised Semantic Point Cloud Segmentation: Towards 10x Fewer Labels (CVPR 2020 · similarity 0.94)
- Weakly Supervised Segmentation on Outdoor 4D Point Clouds With Temporal Matching and Spatial Graph Propagation (CVPR 2022 · similarity 0.92)
- Less Is More: Label Recommendation for Weakly Supervised Point Cloud Semantic Segmentation (AAAI 2024 · similarity 0.92)
- Weakly Supervised Point Cloud Semantic Segmentation via Artificial Oracle (CVPR 2024 · similarity 0.92)
- Multi-Modality Affinity Inference for Weakly Supervised 3D Semantic Segmentation (AAAI 2024 · similarity 0.91)
- LogoSP: Local-global Grouping of Superpoints for Unsupervised Semantic Segmentation of 3D Point Clouds (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.