Research map / Graph learning / GNN architectures
FeaStNet: Feature-Steered Graph Convolutions for 3D Shape Analysis
Read the paperCheck 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 raw XYZ coordinates instead of SHOT features | FeaStNet with translation-invariant assignments · FAUST | accuracy: 63% -> 86% | Helped |
| Add translation-invariant assignments with XYZ input (ablation) | FeaStNet · FAUST | accuracy: 28% -> 86% | Helped |
| Add translation-invariant assignments with SHOT input (ablation) | FeaStNet · FAUST | accuracy: 58% -> 63% | Helped |
| Use a multi-scale architecture instead of single-scale | FeaStNet with XYZ input, without refinement · FAUST | accuracy: 88.1% -> 98.6% | Helped |
| Add refinement (ablation) | FeaStNet multi-scale with XYZ input · FAUST | accuracy: 98.6% -> 98.7% | No clear effect |
| Add refinement (ablation) | FeaStNet single-scale with XYZ input · FAUST | accuracy: 88.1% -> 92.2% | Helped |
| Increase the number of nearest neighbors from 4 to 8 (ablation) | FeaStNet · ShapeNet part | mIoU: 79.9% -> 80.8% | Helped |
| Increase the number of nearest neighbors from 8 to 16 (ablation) | FeaStNet · ShapeNet part | mIoU: 80.8% -> 81.5% | Helped |
| Increase the number of nearest neighbors from 16 to 32 (ablation) | FeaStNet · ShapeNet part | mIoU: 81.5% -> 80.9% | Hurt |
Closest papers by meaning
- Primal-Dual Mesh Convolutional Neural Networks (NeurIPS 2020 · similarity 0.90)
- View-GCN: View-Based Graph Convolutional Network for 3D Shape Analysis (CVPR 2020 · similarity 0.89)
- Learning Local Neighboring Structure for Robust 3D Shape Representation (AAAI 2021 · similarity 0.89)
- Geometric Deep Learning on Graphs and Manifolds Using Mixture Model CNNs (CVPR 2017 · similarity 0.89)
- Geometrically Principled Connections in Graph Neural Networks (CVPR 2020 · similarity 0.88)
- SplineCNN: Fast Geometric Deep Learning With Continuous B-Spline Kernels (CVPR 2018 · 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.