Research map / Label-efficient and robust learning / Recognition and long-tail
Learning Spatial Regularization With Image-Level Supervisions for Multi-Label Image Classification
Check 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 SRN spatial and semantic regularization to ResNet-101 | ResNet-101 · NUS-WIDE | mAP: 59.8 -> 62.0 | Helped |
| Add SRN spatial and semantic regularization to ResNet-101 | ResNet-101 · MS-COCO validation | mAP: 75.2 -> 77.1 | Helped |
| Add SRN spatial and semantic regularization to ResNet-101 | ResNet-101 · WIDER-Attribute | mAP: 85.0 -> 86.2 | Helped |
| Use confidence-weighted attention maps U rather than unweighted maps A (ablation) | ResNet-SRN · NUS-WIDE | mAP: 61.8 -> 62.0 | Helped |
| Use confidence-weighted attention maps U rather than unweighted maps A (ablation) | ResNet-SRN · MS-COCO validation | mAP: 76.1 -> 77.1 | Helped |
| Use confidence-weighted attention maps U rather than unweighted maps A (ablation) | ResNet-SRN · WIDER-Attribute | mAP: 85.4 -> 86.2 | Helped |
| Add two ResNet building blocks to match SRN parameter size | ResNet-101 · NUS-WIDE | mAP: 59.8 -> 59.5 | No clear effect |
| Add two ResNet building blocks to match SRN parameter size | ResNet-101 · MS-COCO validation | mAP: 75.2 -> 75.4 | Helped |
| Add two ResNet building blocks to match SRN parameter size | ResNet-101 · WIDER-Attribute | mAP: 85.0 -> 85.0 | No clear effect |
| Learn semantic relations from initial ResNet label confidences | ResNet-101 · NUS-WIDE | mAP: 59.8 -> 60.1 | No clear effect |
| Learn semantic relations from initial ResNet label confidences | ResNet-101 · MS-COCO validation | mAP: 75.2 -> 75.5 | Helped |
| Remove the top-3 prediction constraint | ResNet-SRN · NUS-WIDE | F1-C: 48.9 -> 58.5 | Helped |
Closest papers by meaning
- Multi-Label Image Recognition With Graph Convolutional Networks (CVPR 2019 · similarity 0.90)
- CNN-RNN: A Unified Framework for Multi-Label Image Classification (CVPR 2016 · similarity 0.90)
- Re-Labeling ImageNet: From Single to Multi-Labels, From Global to Localized Labels (CVPR 2021 · similarity 0.88)
- Regularizing Deep Networks by Modeling and Predicting Label Structure (CVPR 2018 · similarity 0.87)
- Deep Semantic Dictionary Learning for Multi-label Image Classification (AAAI 2021 · similarity 0.87)
- Interactive Multi-Label CNN Learning With Partial Labels (CVPR 2020 · 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.