Research map / Label-efficient and robust learning / Recognition and long-tail
ResLT: Residual Learning for Long-tailed Recognition
Jiequan Cui, Shu Liu, Zhuotao Tian, Zhisheng Zhong, Jiaya Jia
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 residual fusion of nested branches instead of naive summing of independent branches (ablation) | ResLT · CIFAR-10-LT imbalance 0.02 | top-1 accuracy: 61.80 -> 83.46 | Improved |
| use residual fusion instead of vanilla CE training | ResLT · CIFAR-10-LT imbalance 0.02 | top-1 accuracy: 78.50 -> 83.46 | Improved |
| train with ResLT instead of cross-entropy | ResLT · ImageNet-LT ResNeXt-50 few-shot | top-1 accuracy: 13.4 -> 35.5 | Improved |
| ResLT trades many-shot accuracy for medium and few-shot | ResLT · ImageNet-LT ResNeXt-50 many-shot | top-1 accuracy: 67.6 -> 63.0 | Hurt |
| add strong data augmentation (mixup and autoaugment) (ablation) | ResLT · ImageNet-LT ResNeXt-50 | top-1 accuracy: 52.9 -> 56.1 | Improved |
| add strong augmentation and ImageNet pretraining with mixup (ablation) | ResLT · iNaturalist 2018 | top-1 accuracy: 70.2 -> 72.3 | Improved |
| add a third expert (ablation) | ResLT · ImageNet-LT RIDEResNeXt-50 | top-1 accuracy: 56.4 -> 57.6 | Improved |
| add a third expert (ablation) | ResLT · iNaturalist 2018 RIDEResNet-50 | top-1 accuracy: 72.1 -> 72.9 | Improved |
| increase groups from 3 to 4 (ablation) | ResLT · CIFAR-100-LT imbalance 0.01 | top-1 accuracy: 45.34 -> 45.34 | No clear effect |
| increase groups from 3 to 5 (ablation) | ResLT · CIFAR-10-LT imbalance 0.01 | top-1 accuracy: 80.44 -> 80.44 | No clear effect |
| use 2 groups instead of 3 (ablation) | ResLT · CIFAR-10-LT imbalance 0.01 | top-1 accuracy: 80.44 -> 78.83 | Hurt |
| compare ResLT against grouped or separate extra convolutions (ablation) | ResLT · CIFAR-LT | top-1 accuracy | Improved |
Later papers that compared with this method
4 later accepted papers report ResLT as a baseline in their results tables; 3 of them report a better number on most of the results they share with it.
- Feature Fusion from Head to Tail for Long-Tailed Visual Recognition (AAAI 2024 · better on 1 of 1 shared result)
- Dual Compensation Residual Networks for Class Imbalanced Learning (TPAMI 2023 · better on 2 of 2 shared results)
- Local and Global Logit Adjustments for Long-Tailed Learning (ICCV 2023 · better on 0 of 1 shared result)
- Balanced Contrastive Learning for Long-Tailed Visual Recognition (CVPR 2022 · better on 3 of 3 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
- Deep Long-Tailed Learning: A Survey (TPAMI 2023 · similarity 0.90)
- Bag of Tricks for Long-Tailed Visual Recognition with Deep Convolutional Neural Networks (AAAI 2021 · similarity 0.90)
- Feature Fusion from Head to Tail for Long-Tailed Visual Recognition (AAAI 2024 · similarity 0.89)
- DiffuLT: Diffusion for Long-tail Recognition Without External Knowledge (NeurIPS 2024 · similarity 0.88)
- Long-tailed Recognition with Model Rebalancing (NeurIPS 2025 · similarity 0.88)
- SimLTD: Simple Supervised and Semi-Supervised Long-Tailed Object Detection (CVPR 2025 · similarity 0.88)
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.
