Research map / Graph learning / Knowledge and temporal graphs
Graph-Based Global Reasoning Networks
Yunpeng Chen, Marcus Rohrbach, Zhicheng Yan, Yan Shuicheng, Jiashi Feng, Yannis Kalantidis
Check 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 GloRe units at Res4 | ResNet-50 · ImageNet val | Top-1 accuracy: 76.2% -> 78.4% | Improved |
| Add GloRe units at Res3 and Res4 | ResNet-200 · ImageNet val | Top-1 accuracy: 78.3% -> 79.7% | Improved |
| Use 2 graph convolution layers instead of 1 per GloRe unit (ablation) | ResNet-50 + GloRe · ImageNet val | Top-1 accuracy: 77.60% -> 77.62% | No clear effect |
| Use 3 graph convolution layers instead of 1 per GloRe unit (ablation) | ResNet-50 + GloRe · ImageNet val | Top-1 accuracy: 77.60% -> 77.66% | No clear effect |
| Add one GloRe unit after FCN with multi-grid | FCN + multi-grid, ResNet-50 · Cityscapes val | mIoU: 76.45% -> 78.25% | Improved |
| Add a second sequential GloRe unit (ablation) | FCN + multi-grid, ResNet-50 · Cityscapes val | mIoU: 78.25% -> 77.84% | Hurt |
| Add GloRe units | ResNet-50 (3D) · Kinetics-400 val | Clip Top-1: 67.12% -> 68.02% | Improved |
| Add GloRe units | ResNet-101 (3D) · Kinetics-400 val | Video Top-1: 75.69 % -> 76.09% | Improved |
| Insert the single GloRe unit at Res4 rather than Res3 (ablation) | ResNet-50 · ImageNet val | Top-1 accuracy | Improved |
| Add GloRe units across more stages rather than all at Res4 (ablation) | ResNet-50 · ImageNet val | Top-1 accuracy | Hurt |
| Add three GloRe units at Res4 (ablation) | ResNet-50 · ImageNet val | Top-1 accuracy: 78.4% | Improved |
Closest papers by meaning
- Region-Based Global Reasoning Networks (AAAI 2020 · similarity 0.90)
- Symbolic Graph Reasoning Meets Convolutions (NeurIPS 2018 · similarity 0.88)
- Iterative Visual Reasoning Beyond Convolutions (CVPR 2018 · similarity 0.88)
- Spatial-Aware Graph Relation Network for Large-Scale Object Detection (CVPR 2019 · similarity 0.87)
- GI-GCN: Global Interacted Graph Convolutional Networks via Dominant Sets for Graph Classification (ICML 2026 · similarity 0.87)
- GITA: Graph to Visual and Textual Integration for Vision-Language Graph Reasoning (NeurIPS 2024 · similarity 0.86)
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.
