Research map / Point clouds / Segmentation and detection
Don’t Starve the Boundaries: Boundary-Constrained Label Propagation for Weakly Supervised 3D Segmentation
Shuwei Wu, shuo jin, Zhijin He, Siyue Yu, ENG LIM, Qiufeng Wang, Jimin Xiao
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 consistency framework to sparse-label MinkowskiNet (ablation) | MinkowskiNet · S3DIS Area 5 0.01% | mIoU: 49.3 -> 55.8 | Improved |
| add first-round boundary-constrained label propagation (BCLP1) (ablation) | consistency baseline · S3DIS Area 5 0.01% | mIoU: 55.8 -> 56.7 | Improved |
| add BAR-triggered iterative propagation (BCLPn) (ablation) | consistency baseline · S3DIS Area 5 0.01% | mIoU: 56.7 -> 60.7 | Improved |
| add position-aware divide-and-conquer (PADC) (ablation) | Bound3D · S3DIS Area 5 0.01% | B-mIoU: 52.9 -> 54.5 | Improved |
| use distinct boundary/core thresholds (ablation) | BCLPn · S3DIS Area 5 0.01% | mIoU: 58.9 -> 61.2 | Improved |
| add farthest angle sampling (ablation) | BCLPn · S3DIS Area 5 0.01% | mIoU: 59.3 -> 61.2 | Improved |
| add distance weighting (ablation) | BCLPn · S3DIS Area 5 0.01% | mIoU: 60.0 -> 61.2 | Improved |
| use stop-gradient in PADC auxiliary head (ablation) | Bound3D · S3DIS Area 5 0.01% | mIoU: 60.3 -> 61.2 | Improved |
| increase boundary consistency loss weight to 1.25x (ablation) | consistency baseline · S3DIS | mIoU and B-mIoU | Hurt |
| BCLP pseudo-labels instead of SAM-based pseudo-labels | pseudo-label generation · S3DIS 1 pt/obj | pseudo-label accuracy: 77.1 -> 94.4 | Improved |
| Bound3D-PTv2 vs AO-PTv2 under same backbone | Bound3D · S3DIS 1 pt/obj | mIoU: 62.7 -> 63.5 | Improved |
| Bound3D-MinkNet vs You et al.-PTv3 | Bound3D · S3DIS 1 pt/obj | mIoU: 64.4 -> 61.9 | Hurt |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- CPCM — CPCM: Contextual Point Cloud Modeling for Weakly-supervised Point Cloud Semantic Segmentation (ICCV 2023) · better on 1 of 1 shared result
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- Weakly Supervised 3D Segmentation via Receptive-Driven Pseudo Label Consistency and Structural Consistency (AAAI 2023 · similarity 0.90)
- Weakly Supervised Point Cloud Semantic Segmentation via Artificial Oracle (CVPR 2024 · similarity 0.90)
- Less Is More: Label Recommendation for Weakly Supervised Point Cloud Semantic Segmentation (AAAI 2024 · similarity 0.89)
- Fully Data-Driven Pseudo Label Estimation for Pointly-Supervised Panoptic Segmentation (AAAI 2024 · similarity 0.89)
- W2P: Switching from Weak Supervision to Partial Supervision for Semantic Segmentation (AAAI 2024 · similarity 0.89)
- LogoSP: Local-global Grouping of Superpoints for Unsupervised Semantic Segmentation of 3D Point Clouds (CVPR 2025 · 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.
