Research map / Point clouds / Segmentation and detection
Hierarchical Point-Edge Interaction Network for Point Cloud Semantic Segmentation
Li Jiang, Hengshuang Zhao, Shu Liu, Xiaoyong Shen, Chi-Wing Fu, Jiaya Jia
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 |
|---|---|---|---|
| use concatenation as edge function instead of subtraction (ablation) | Point-Edge network · ScanNet v2 val | mIoU: 58.31 -> 63.36 | Improved |
| use concatenation as edge function instead of summation (ablation) | Point-Edge network · ScanNet v2 val | mIoU: 57.86 -> 63.36 | Improved |
| use concatenation as edge function instead of Hadamard product (ablation) | Point-Edge network · ScanNet v2 val | mIoU: 59.07 -> 63.36 | Improved |
| use plain concatenation instead of ConcatSub edge function (ablation) | Point-Edge network · ScanNet v2 val | mIoU: 63.09 -> 63.36 | No clear effect |
| use concatenation as edge function instead of subtraction (ablation) | Point-Edge network · S3DIS Area 5 | mIoU: 58.85 -> 61.85 | Improved |
| use concatenation as edge function instead of ConcatSub (ablation) | Point-Edge network · S3DIS Area 5 | mIoU: 59.37 -> 61.85 | Improved |
| use max-pooling plus concatenation for message passing instead of AdaAggre with softmax (ablation) | Point-Edge network · ScanNet v2 val | mIoU: 56.44 -> 63.36 | Improved |
| use max-pooling plus concatenation instead of AdaAggre without softmax (ablation) | Point-Edge network · ScanNet v2 val | mIoU: 55.01 -> 63.36 | Improved |
| add softmax to adaptive aggregation weights (ablation) | AdaAggre message passing · ScanNet v2 val | mIoU: 55.01 -> 56.44 | Improved |
| add hierarchical graph construction with edge upsampling (ablation) | Point-Edge network · ScanNet v2 val | mIoU: 57.01 -> 63.36 | Improved |
| add hierarchical graph construction with edge upsampling (ablation) | Point-Edge network · ScanNet v2 val | mAcc: 66.52 -> 72.61 | Improved |
| add hierarchical graph construction with edge upsampling (ablation) | Point-Edge network · ScanNet v2 val | OA: 83.57 -> 86.13 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 2 of 2 on most shared results.
- PointCNN — PointCNN: Convolution On X-Transformed Points (NeurIPS 2018) · better on 5 of 5 shared results
- PointConv — PointConv: Deep Convolutional Networks on 3D Point Clouds (CVPR 2019) · 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
- Exploiting Local and Global Structure for Point Cloud Semantic Segmentation with Contextual Point Representations (NeurIPS 2019 · similarity 0.91)
- Multi-Path Region Mining for Weakly Supervised 3D Semantic Segmentation on Point Clouds (CVPR 2020 · similarity 0.90)
- Geometry and Uncertainty-Aware 3D Point Cloud Class-Incremental Semantic Segmentation (CVPR 2023 · similarity 0.90)
- SSPC-Net: Semi-supervised Semantic 3D Point Cloud Segmentation Network (AAAI 2021 · similarity 0.89)
- Less Is More: Label Recommendation for Weakly Supervised Point Cloud Semantic Segmentation (AAAI 2024 · similarity 0.88)
- Weakly Supervised Point Cloud Semantic Segmentation via Artificial Oracle (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.
