Research map / Label-efficient and robust learning / Recognition and long-tail
Dual Compensation Residual Networks for Class Imbalanced Learning
Ruibing Hou, Hong Chang, Bingpeng Ma, Shiguang Shan, Xilin Chen
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 |
|---|---|---|---|
| use DCRNets instead of ResLT residual mechanism | ResNet-10 backbone · ImageNet-LT (>180 epochs) | top-1 accuracy (%): 43.8 -> 45.0 | Improved |
| use DCRNets instead of ResLT residual mechanism | ResNeXt-50 backbone · ImageNet-LT (>180 epochs) | top-1 accuracy (%): 52.9 -> 55.0 | Improved |
| use DCRNets instead of SSP self-supervised initialization | ResNet-50 backbone · ImageNet-LT | top-1 accuracy (%): 51.3 -> 54.1 | Improved |
| use DCRNets instead of decoupled DisAlign | ResNet-50 backbone · ImageNet-LT (90 epochs) | top-1 accuracy (%): 51.3 -> 53.2 | Improved |
| add DCRNets modules to RIDE multi-expert network (2 experts) | RIDE ResNet-50 · ImageNet-LT | top-1 accuracy (%): 54.4 -> 55.7 | Improved |
| add DCRNets modules to RIDE multi-expert network (3 experts) | RIDE ResNet-50 · ImageNet-LT | top-1 accuracy (%): 54.9 -> 56.9 | Improved |
| use DCRNets instead of Hybrid-PSC | ResNet-50 · iNaturalist18 (90 epochs) | top-1 accuracy (%): 68.1 -> 70.3 | Improved |
| add DCRNets modules to RIDE (2 experts) | RIDE ResNet-32 · CIFAR100-LT | top-1 accuracy (%): 47.0 -> 55.0 | Improved |
| train longer (90 to >180 epochs) | DCRNets ResNet-50 · ImageNet-LT | top-1 accuracy (%): 53.2 -> 54.1 | Improved |
Earlier methods this paper compares with
Named baselines from earlier accepted papers in this paper's results tables: better than 1 of 1 on most shared results.
- ResLT — ResLT: Residual Learning for Long-tailed Recognition (TPAMI 2022) · better on 2 of 2 shared results
Matched by method name in results tables read by AI. Authors' own tables tend to favour the newer method, and a comparison can be missed when a method is renamed.
Closest papers by meaning
- A Tale of Two Classes: Adapting Supervised Contrastive Learning to Binary Imbalanced Datasets (CVPR 2025 · similarity 0.84)
- Inverse Weight-Balancing for Deep Long-Tailed Learning (AAAI 2024 · similarity 0.83)
- Balancing the Scales: A Theoretical and Algorithmic Framework for Learning from Imbalanced Data (ICML 2025 · similarity 0.83)
- BaCon: Boosting Imbalanced Semi-supervised Learning via Balanced Feature-Level Contrastive Learning (AAAI 2024 · similarity 0.83)
- Long-tailed Recognition with Model Rebalancing (NeurIPS 2025 · similarity 0.83)
- Difficulty-aware Balancing Margin Loss for Long-tailed Recognition (AAAI 2025 · similarity 0.83)
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.
