Research map / Point clouds / Segmentation and detection
SQN: Weakly-Supervised Semantic Segmentation of Large-Scale 3D Point Clouds
Qingyong Hu, Bo Yang, Guangchi Fang, Yulan Guo, Aleš Leonardis, Niki Trigoni, Andrew Markham
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 semantic query to RandLA-Net under 0.1% random labels | RandLA-Net · S3DIS Area-5 | mIoU: 52.9 -> 61.4 | Improved |
| query features at all four encoder layers instead of only the first (ablation) | SQN (0.1% labels) · S3DIS Area-5 | mIoU: 22.89 -> 61.41 | Improved |
| query features at all four encoder layers instead of only the last (ablation) | SQN (0.1% labels) · S3DIS Area-5 | mIoU: 46.02 -> 61.41 | Improved |
| add the 4th-layer query to layers 1-3 (ablation) | SQN (0.1% labels) · S3DIS Area-5 | mIoU: 54.21 -> 61.41 | Improved |
| replace PointNet++ propagation layers with semantic query decoder (ablation) | SQN (0.1% labels) · S3DIS Area-5 | mIoU: 53.4 -> 61.4 | Improved |
| replace RandLA-Net propagation layers with semantic query decoder (ablation) | SQN (0.1% labels) · ScanNet val | mIoU: 51.2 -> 58.4 | Improved |
| replace RandLA-Net propagation layers with semantic query decoder (ablation) | SQN (0.1% labels) · Toronto3D | mIoU: 71.7 -> 77.7 | Improved |
| vary number of queried neighbours from 1 to 25 (ablation) | SQN · S3DIS Area-5 | mIoU | No clear effect |
| resample which 0.1% of points are labelled over 5 runs (ablation) | SQN · S3DIS Area-5 | mIoU: 60.42 mean, 0.93 STD | No clear effect |
| reduce labels from 0.1% to 0.01% (ablation) | SQN · S3DIS Area-5 | mIoU: 61.41 -> 45.30 | Hurt |
| increase labels from 0.1% to 10% (ablation) | SQN · S3DIS Area-5 | mIoU: 61.41 -> 64.67 | Improved |
| add SQN decoder to MinkUNet at 0.1% labels | MinkUNet · SemanticKITTI val | mIoU: 55.5 -> 55.8 | No clear effect |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 0 of 1 on most shared results.
- RandLA-Net (100%) — RandLA-Net: Efficient Semantic Segmentation of Large-Scale Point Clouds (CVPR 2020) · better on 2 of 10 shared results
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 Semantic Point Cloud Segmentation: Towards 10x Fewer Labels (CVPR 2020 · similarity 0.94)
- Multi-Path Region Mining for Weakly Supervised 3D Semantic Segmentation on Point Clouds (CVPR 2020 · similarity 0.93)
- Weakly Supervised Point Cloud Semantic Segmentation via Artificial Oracle (CVPR 2024 · similarity 0.92)
- Less Is More: Label Recommendation for Weakly Supervised Point Cloud Semantic Segmentation (AAAI 2024 · similarity 0.91)
- Multi-Modality Affinity Inference for Weakly Supervised 3D Semantic Segmentation (AAAI 2024 · similarity 0.90)
- Point Cloud Semantic Segmentation with Sparse and Inhomogeneous Annotations (AAAI 2025 · similarity 0.90)
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.
