Research map / Label-efficient and robust learning / Recognition and long-tail
Long-Tailed Multi-Label Visual Recognition by Collaborative Training on Uniform and Re-Balanced Samplings
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 |
|---|---|---|---|
| Train collaboratively on uniform and re-balanced samples with cross-branch logit consistency and learnable logit compensation | ResNet50 multi-label recognition · VOC-LT | mAP: 78.42 -> 81.44 | Helped |
| Train collaboratively on uniform and re-balanced samples with cross-branch logit consistency and learnable logit compensation | ResNet50 multi-label recognition · COCO-LT | mAP: 54.33 -> 56.90 | Helped |
| Use both uniform and re-balanced branches rather than only the uniform branch (ablation) | Long-tailed multi-label recognition · VOC-LT | mAP: 77.15 -> 79.42 | Helped |
| Use both uniform and re-balanced branches rather than only the uniform branch (ablation) | Long-tailed multi-label recognition · COCO-LT | mAP: 53.15 -> 54.71 | Helped |
| Add cross-branch logit consistency to the two-branch ensemble (ablation) | Long-tailed multi-label recognition · VOC-LT | mAP: 79.42 -> 81.22 | Helped |
| Add cross-branch logit consistency to the two-branch ensemble (ablation) | Long-tailed multi-label recognition · COCO-LT | mAP: 54.71 -> 56.62 | Helped |
| Add learnable logit compensation (ablation) | Two-branch model with logit consistency · VOC-LT | mAP: 81.22 -> 81.44 | Helped |
| Add learnable logit compensation (ablation) | Two-branch model with logit consistency · COCO-LT | mAP: 56.62 -> 56.90 | Helped |
| Add horizontal-flip test-time augmentation (ablation) | Proposed model · VOC-LT | mAP: 81.44 -> 81.79 | Helped |
| Add horizontal-flip test-time augmentation (ablation) | Proposed model · COCO-LT | mAP: 56.90 -> 57.28 | Helped |
| Use sigmoid-probability consistency instead of logit consistency (ablation) | Two-branch model · VOC-LT | mAP: 81.44 -> 80.32 | Hurt |
| Use sigmoid-probability consistency instead of logit consistency (ablation) | Two-branch model · COCO-LT | mAP: 56.90 -> 56.03 | Hurt |
Closest papers by meaning
- Global and Local Mixture Consistency Cumulative Learning for Long-Tailed Visual Recognitions (CVPR 2023 · similarity 0.91)
- MLC-NC: Long-Tailed Multi-Label Image Classification Through the Lens of Neural Collapse (AAAI 2025 · similarity 0.89)
- Long-tailed Recognition with Model Rebalancing (NeurIPS 2025 · similarity 0.89)
- Decoupled Optimisation for Long-Tailed Visual Recognition (AAAI 2024 · similarity 0.88)
- Feature Fusion from Head to Tail for Long-Tailed Visual Recognition (AAAI 2024 · similarity 0.88)
- Robust Visual Recognition with Class-Imbalanced Open-World Noisy Data (AAAI 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.