Research map / Graph learning / GNN architectures
Dynamic Graph Message Passing Networks for Visual Recognition
Li Zhang, Mohan Chen, Anurag Arnab, Xiangyang Xue, Philip H. S. Torr
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 dynamic affinity (ablation) | Dilated FCN (ResNet-101) · Cityscapes val (single scale) | mIoU: 75.0 -> 76.5 | Improved |
| add dynamic weights (ablation) | DGMN w/ DA · Cityscapes val (single scale) | mIoU: 76.5 -> 79.1 | Improved |
| add dynamic sampling (ablation) | DGMN w/ DA+DW · Cityscapes val (single scale) | mIoU: 79.1 -> 80.4 | Improved |
| use DGMN instead of Non-local | Dilated FCN (ResNet-101) · Cityscapes val (single scale) | mIoU: 79.0 -> 80.4 | Improved |
| use DGMN instead of Non-local on Mask R-CNN | Mask R-CNN R50 · COCO val | AP box: 39.0 -> 39.5 | Improved |
| add dynamic sampling and dynamic weights/affinities to deformable message passing (ablation) | Mask R-CNN R50 · COCO val | AP box: 38.7 -> 39.5 | Improved |
| insert DGMN (C5) into Mask R-CNN | Mask R-CNN R50 · COCO val | AP box: 37.8 -> 40.2 | Improved |
| insert DGMN in C4 and C5 instead of C5 only (ablation) | Mask R-CNN R50 · COCO test-dev | AP box: 40.2 -> 41.0 | Improved |
| add DGMN (C5) to ResNet-101 Mask R-CNN | Mask R-CNN R101 · COCO test-dev | AP box: 40.2 -> 41.9 | Improved |
| add DGMN (C5) to ResNeXt-101 Mask R-CNN | Mask R-CNN X101 · COCO test-dev | AP box: 42.6 -> 44.3 | Improved |
| use DGMN2-Small backbone instead of ResNet-50 in Sparse R-CNN | Sparse R-CNN · COCO val | AP: 45.0 -> 48.2 | Improved |
| use DGMN2-Tiny instead of PVT-Tiny | backbone · ImageNet-1k val | top-1 accuracy: 75.1 -> 78.7 | Improved |
Closest papers by meaning
- Dynamic Graph Message Passing Networks (CVPR 2020 · similarity 0.96)
- A Survey on Graph Neural Networks and Graph Transformers in Computer Vision: A Task-Oriented Perspective (TPAMI 2024 · similarity 0.89)
- Attention-Driven Dynamic Graph Convolutional Network for Multi-Label Image Recognition (ECCV 2020 · similarity 0.89)
- Towards Dynamic Message Passing on Graphs (NeurIPS 2024 · similarity 0.89)
- Class-wise Dynamic Graph Convolution for Semantic Segmentation (ECCV 2020 · similarity 0.88)
- Graph Convolutions Enrich the Self-Attention in Transformers! (NeurIPS 2024 · similarity 0.87)
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.
