Research map / Label-efficient and robust learning / Recognition and long-tail
Transformer-Based Dual Relation Graph for Multi-Label Image Recognition
Jiawei Zhao, Ke Yan, Yifan Zhao, Xiaowei Guo, Feiyue Huang, Jia Li
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 Transformer structural relation (ablation) | ResNet-101 baseline · MS-COCO | mAP: 78.6 -> 82.9 | Improved |
| add Transformer structural relation (ablation) | ResNet-101 baseline · VOC 2007 | mAP: 92.9 -> 94.3 | Improved |
| add cross-scale attention (ablation) | Transformer structural relation · MS-COCO | mAP: 82.9 -> 83.9 | Improved |
| add GCN semantic relation (ablation) | ResNet-101 baseline · MS-COCO | mAP: 78.6 -> 82.5 | Improved |
| add semantic-aware constraints (ablation) | GCN semantic relation · MS-COCO | mAP: 82.5 -> 83.5 | Improved |
| add Transformer to GCN (ablation) | GCN semantic relation · VOC 2007 | mAP: 93.4 -> 94.6 | Improved |
| add 1/64 scale features (ablation) | Transformer structural relation · MS-COCO | mAP: 82.9 -> 83.1 | Improved |
| add 1/16 scale features (ablation) | Transformer structural relation · MS-COCO | mAP: 83.1 -> 83.3 | Improved |
| multiplication instead of summation in cross attention (ablation) | cross-scale attention · MS-COCO | mAP: 83.2 -> 83.9 | Improved |
| transformer units instead of MLP in cross-scale module (ablation) | cross-scale attention · MS-COCO | mAP: 83.3 -> 83.9 | Improved |
| learnable correlation matrix instead of static (ablation) | semantic relation graph · MS-COCO | mAP: 81.9 -> 82.5 | Improved |
| add structural guidance to joint correlation matrix (ablation) | semantic relation graph · MS-COCO | mAP: 82.5 -> 83.0 | Improved |
Closest papers by meaning
- Learning Semantic-Specific Graph Representation for Multi-Label Image Recognition (ICCV 2019 · similarity 0.91)
- Improving Multi-label Learning with Missing Labels by Structured Semantic Correlations (ECCV 2016 · similarity 0.89)
- Modular Graph Transformer Networks for Multi-Label Image Classification (AAAI 2021 · similarity 0.89)
- Relationformer: A Unified Framework for Image-to-Graph Generation (ECCV 2022 · similarity 0.88)
- Multi-Label Image Recognition With Graph Convolutional Networks (CVPR 2019 · similarity 0.88)
- Scene-Aware Label Graph Learning for Multi-Label Image Classification (ICCV 2023 · 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.
