Research map / Graph learning / Knowledge and temporal graphs
Vision GNN: An Image is Worth Graph of Nodes
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 |
|---|---|---|---|
| add FC layers in Grapher and FFN module to plain graph convolution (ablation) | ViG-Ti · ImageNet-1k | top-1 accuracy: 67.0 -> 73.9 | Helped |
| add FFN module to GraphConv+FC (ablation) | ViG-Ti · ImageNet-1k | top-1 accuracy: 73.4 -> 73.9 | Helped |
| add FC in Grapher to GraphConv+FFN (ablation) | ViG-Ti · ImageNet-1k | top-1 accuracy: 73.6 -> 73.9 | Helped |
| use EdgeConv instead of Max-Relative GraphConv (ablation) | ViG-Ti · ImageNet-1k | top-1 accuracy: 73.9 -> 74.3 (FLOPs 1.3B -> 2.4B) | Helped |
| use GIN instead of Max-Relative GraphConv (ablation) | ViG-Ti · ImageNet-1k | top-1 accuracy: 73.9 -> 72.8 | Hurt |
| increase neighbors K from 3 to 9 (ablation) | ViG-Ti · ImageNet-1k | top-1 accuracy: 72.2 -> 73.6 | Helped |
| increase neighbors K from 15 to 20 (ablation) | ViG-Ti · ImageNet-1k | top-1 accuracy: 73.5 -> 73.3 | Hurt |
| linearly increase K from 9 to 18 with depth instead of fixed K (ablation) | ViG-Ti · ImageNet-1k | top-1 accuracy: 73.6 -> 73.9 | Helped |
| increase heads h from 1 to 8 (ablation) | ViG-Ti · ImageNet-1k | top-1 accuracy: 74.2 -> 73.7 (1.6B -> 1.2B FLOPs) | No clear effect |
| graph-based ViG backbone instead of transformer | isotropic vision backbone · ImageNet-1k | top-1 accuracy: 72.2 -> 73.9 | Helped |
| Pyramid ViG-S backbone instead of Swin-T | RetinaNet · COCO val2017 | mAP: 41.5 -> 41.8 | Helped |
Closest papers by meaning
- GreedyViG: Dynamic Axial Graph Construction for Efficient Vision GNNs (CVPR 2024 · similarity 0.92)
- Vision Graph Prompting via Semantic Low-Rank Decomposition (ICML 2025 · similarity 0.88)
- Hypergraph Vision Transformers: Images are More than Nodes, More than Edges (CVPR 2025 · similarity 0.87)
- DVHGNN: Multi-Scale Dilated Vision HGNN for Efficient Vision Recognition (CVPR 2025 · similarity 0.86)
- Generalizing CNNs to graphs with learnable neighborhood quantization (NeurIPS 2024 · similarity 0.86)
- Open Your Eyes: Vision Enhances Message Passing Neural Networks in Link Prediction (ICML 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.