Research map / Label-efficient and robust learning / Recognition and long-tail
Constructing Balance from Imbalance for Long-Tailed Image Recognition
Yue Xu, Yong-Lu Li, Jiefeng Li, Cewu Lu
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 DLSA label-space adjustment | PaCo+BalSoftmax · ImageNet-LT | overall accuracy (%): 54.9 -> 56.3 | Improved |
| add DLSA label-space adjustment | CE+BalSoftmax · ImageNet-LT | overall accuracy (%): 42.8 -> 43.9 | Improved |
| add DLSA label-space adjustment | PaCo+BalSoftmax · Places-LT | overall accuracy (%): 40.9 -> 42.1 | Improved |
| add DLSA label-space adjustment | PaCo+BalSoftmax · iNaturalist18 | overall accuracy (%): 71.8 -> 72.8 | Improved |
| add weighted MLE loss (ablation) | DLSA · ImageNet-LT | overall accuracy (%): 54.7 -> 56.3 | Improved |
| add cluster balancedness loss (ablation) | DLSA · ImageNet-LT | overall accuracy (%): 55.2 -> 56.3 | Improved |
| add purity loss (ablation) | DLSA · ImageNet-LT | overall accuracy (%): 55.4 -> 56.3 | Improved |
| add weighted MLE loss (ablation) | DLSA · Places-LT | overall accuracy (%): 40.8 -> 42.1 | Improved |
| use 500 clusters instead of 300 (ablation) | DLSA · ImageNet-LT | overall accuracy (%): 55.5 -> 56.3 | Improved |
| use 500 clusters instead of 1000 (ablation) | DLSA · ImageNet-LT | overall accuracy (%): 55.1 -> 56.3 | Improved |
| use 3 flow filter blocks instead of 2 (ablation) | DLSA · ImageNet-LT | overall accuracy (%): 55.4 -> 56.3 | Improved |
| use 3 flow filter blocks instead of 4 (ablation) | DLSA · ImageNet-LT | overall accuracy (%): 55.8 -> 56.3 | Improved |
Closest papers by meaning
- Long-tailed Recognition with Model Rebalancing (NeurIPS 2025 · similarity 0.93)
- Difficulty-aware Balancing Margin Loss for Long-tailed Recognition (AAAI 2025 · similarity 0.92)
- Long-Tailed Recognition via Information-Preservable Two-Stage Learning (NeurIPS 2025 · similarity 0.91)
- Exploring Weight Balancing on Long-Tailed Recognition Problem (ICLR 2024 · similarity 0.90)
- MLC-NC: Long-Tailed Multi-Label Image Classification Through the Lens of Neural Collapse (AAAI 2025 · similarity 0.90)
- Feature Fusion from Head to Tail for Long-Tailed Visual Recognition (AAAI 2024 · similarity 0.90)
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.
