Research map / Label-efficient and robust learning / Recognition and long-tail
Adaptive Logit Adjustment Loss for Long-Tailed Visual Recognition
Yan Zhao, Weicong Chen, Xu Tan, Kai Huang, Jihong Zhu
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 |
|---|---|---|---|
| Replace DisAlign with ALA Loss | ResNet-50 · ImageNet-LT test | Top-1 accuracy (All): 51.3 -> 52.4 | Improved |
| Replace DisAlign with ALA Loss | ResNeXt-50 · ImageNet-LT test | Top-1 accuracy (All): 52.6 -> 53.3 | Improved |
| Replace LADE with ALA Loss | ResNet-50 · iNaturalist 2018 validation | Top-1 accuracy (All): 70.0 -> 70.7 | Improved |
| Replace DisAlign with ALA Loss | ResNet-152 · Places-LT test | Top-1 accuracy (All): 39.3 -> 40.1 | Improved |
| Add a difficulty factor to Cross Entropy (ablation) | ImageNet-LT classification with ResNeXt-50 · ImageNet-LT test, Many-shot | Top-1 accuracy: 65.9 -> 64.7 | Hurt |
| Add a difficulty factor to Cross Entropy (ablation) | ImageNet-LT classification with ResNeXt-50 · ImageNet-LT test, Few-shot | Top-1 accuracy: 7.7 -> 28.6 | Improved |
| Add a quantity factor to Cross Entropy (ablation) | ImageNet-LT classification with ResNeXt-50 · ImageNet-LT test, Few-shot | Top-1 accuracy: 7.7 -> 33.0 | Improved |
| Replace LDAM's quantity adjustment with the proposed quantity factor (ablation) | ImageNet-LT classification with ResNeXt-50 · ImageNet-LT test, Few-shot | Top-1 accuracy: 31.9 -> 33.0 | Improved |
| Combine the difficulty factor with LDAM's adjustment (ablation) | ImageNet-LT classification with ResNeXt-50 · ImageNet-LT test, All | Top-1 accuracy: 51.3 -> 52.0 | Improved |
| Combine the difficulty factor with the proposed quantity factor (ablation) | ImageNet-LT classification with ResNeXt-50 · ImageNet-LT test, All | Top-1 accuracy: 51.5 -> 53.3 | Improved |
Closest papers by meaning
- Rethinking Classifier Re-Training in Long-Tailed Recognition: Label Over-Smooth Can Balance (ICLR 2025 · similarity 0.89)
- Key Point Sensitive Loss for Long-tailed Visual Recognition (TPAMI 2022 · similarity 0.89)
- Long-Tailed Visual Recognition via Gaussian Clouded Logit Adjustment (CVPR 2022 · similarity 0.88)
- Multiplicative Logit Adjustment Approximates Neural-Collapse-Aware Decision Boundary Adjustment (ICLR 2025 · similarity 0.87)
- Decoupled Optimisation for Long-Tailed Visual Recognition (AAAI 2024 · similarity 0.87)
- Feature Fusion from Head to Tail for Long-Tailed Visual Recognition (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.
