Research map / Point clouds / Point cloud classification
TopoGAT: Plug-and-Play Topological Graph Attention for Fine-Grained 3D Segmentation
Xinyu Jiang, Lech Szymanski, Steven Mills
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 TopoGAT | PointNet++ · ShapeNetPart | Instance mIoU: 84.8 -> 85.8 | Improved |
| add TopoGAT | Point-MAE · ShapeNetPart | Instance mIoU: 86.1 -> 86.7 | Improved |
| add TopoGAT | PointNet++ · S3DIS (xyz only) | Class avg mIoU: 52.5 -> 55.8 | Improved |
| add TopoGAT | Point-MAE · S3DIS (xyz only) | Class avg mIoU: 60.4 -> 61.9 | Improved |
| add TopoGAT | TopoPointNet++ · ShapeNetPart | Class mIoU: 82.8 -> 82.8 | No clear effect |
| add naive topology embedding concatenation (ablation) | PointNet++ and DGCNN · ShapeNetPart | Instance mIoU: 84.8 -> 84.7 | No clear effect |
| add FiLM attention regression (ablation) | PointNet++ · ShapeNetPart | Instance mIoU: 84.8 -> 85.6 | Improved |
| add FiLM attention regression (ablation) | DGCNN · ShapeNetPart | Instance mIoU: 85.0 -> 85.8 | Improved |
| add SE block alone (ablation) | PointNet++ and DGCNN · ShapeNetPart | Instance mIoU: 84.8 -> 84.9 | Improved |
| combine all components (full TopoGAT) over FiLM only (ablation) | PointNet++ · ShapeNetPart | Instance mIoU: 85.6 -> 85.8 | Improved |
| increase KNN k from 20 to 35 (ablation) | PointNet++-TopoGAT · ShapeNetPart | Instance mIoU: 85.4 -> 85.8 | Improved |
| increase KNN k from 20 to 30 (ablation) | DGCNN-TopoGAT · ShapeNetPart | Instance mIoU: 85.5 -> 85.9 | Improved |
Closest papers by meaning
- Persistent Homology Based Graph Convolution Network for Fine-Grained 3D Shape Segmentation (ICCV 2021 · similarity 0.89)
- SegGraph: Leveraging Graphs of SAM Segments for Few-Shot 3D Part Segmentation (NeurIPS 2025 · similarity 0.87)
- TopoTTA: Topology-Enhanced Test-Time Adaptation for Tubular Structure Segmentation (ICCV 2025 · similarity 0.87)
- Learning Fine-Grained Segmentation of 3D Shapes Without Part Labels (CVPR 2021 · similarity 0.86)
- Topograph: An Efficient Graph-Based Framework for Strictly Topology Preserving Image Segmentation (ICLR 2025 · similarity 0.86)
- TopoGaussian: Inferring Internal Topology Structures from Visual Clues (ICLR 2025 · similarity 0.85)
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.
