mlsanityData to 7 Oct 2026

Decoupling Representation and Classifier for Long-Tailed Recognition

arXiv 1910.09217 ↗First posted 21 Oct 2019cs.CVcited by 806

Not on the map. Why?

Cited by 21 in Diffusion models, but placed by its own abstract, which didn’t fit.

Abstract (arXiv)

The long-tail distribution of the visual world poses great challenges for deep learning based classification models on how to handle the class imbalance problem. Existing solutions usually involve class-balancing strategies, e.g., by loss re-weighting, data re-sampling, or transfer learning from head- to tail-classes, but most of them adhere to the scheme of jointly learning representations and classifiers. In this work, we decouple the learning procedure into representation learning and classification, and systematically explore how different balancing strategies affect them for long-tailed recognition. The findings are surprising: (1) data imbalance might not be an issue in learning high-quality representations; (2) with representations learned with the simplest instance-balanced (natural) sampling, it is also possible to achieve strong long-tailed recognition ability by adjusting only the classifier. We conduct extensive experiments and set new state-of-the-art performance on common long-tailed benchmarks like ImageNet-LT, Places-LT and iNaturalist, showing that it is possible to outperform carefully designed losses, sampling strategies, even complex modules with memory, by using a straightforward approach that decouples representation and classification. Our code is available at https://github.com/facebookresearch/classifier-balancing.

Key facts. Cited by 806 papers in the map (data through 7 Oct 2026). First posted on arXiv 21 Oct 2019.

Builds on

No reference list yet.

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2021: 128; 2022: 166; 2023: 164; 2024: 151; 2025: 90; 2026: 107.

806 papers.

Sort:

By area:

  1. 2021Deep Long-Tailed Learning: A Surveycited by 416
  2. 2022Last Layer Re-Training is Sufficient for Robustness to Spurious Correlationscited by 379
  3. 2021DER: Dynamically Expandable Representation for Class Incremental Learningcited by 355
  4. 2022Language-Grounded Indoor 3D Semantic Segmentation in the Wildcited by 243
  5. 2021No Fear of Heterogeneity: Classifier Calibration for Federated Learning with Non-IID Datacited by 241
  6. 2022FOSTER: Feature Boosting and Compression for Class-Incremental Learningcited by 228
  7. 2021Parametric Contrastive Learningcited by 217
  8. 2021Improving Calibration for Long-Tailed Recognitioncited by 212
  9. 2021Distribution Alignment: A Unified Framework for Long-tail Visual Recognitioncited by 198
  10. 2021FedBABU: Towards Enhanced Representation for Federated Image Classificationcited by 180
  11. 2021CReST: A Class-Rebalancing Self-Training Framework for Imbalanced Semi-Supervised Learningcited by 174
  12. 2021Open-World Semi-Supervised Learningcited by 165
38 more
  1. 2021Delving into Deep Imbalanced Regressioncited by 157
  2. 2022On Feature Learning in the Presence of Spurious Correlationscited by 145
  3. 2022Balanced Contrastive Learning for Long-Tailed Visual Recognitioncited by 143
  4. 2021Bipartite Graph Network with Adaptive Message Passing for Unbiased Scene Graph Generationcited by 137
  5. 2021Targeted Supervised Contrastive Learning for Long-Tailed Recognitioncited by 132
  6. 2021Contrastive Learning based Hybrid Networks for Long-Tailed Image Classificationcited by 130
  7. 2021Self-supervised Learning is More Robust to Dataset Imbalancecited by 122
  8. 2022Diffusion Probabilistic Model Made Slimcited by 121
  9. 2021Instance-Conditioned GANcited by 119
  10. 2023Change is Hard: A Closer Look at Subpopulation Shiftcited by 119
  11. 2021Overcoming Catastrophic Forgetting in Incremental Few-Shot Learning by Finding Flat Minimacited by 118
  12. 2022Federated Learning with Label Distribution Skew via Logits Calibrationcited by 108
  13. 2022Long-Tailed Recognition via Weight Balancingcited by 107
  14. 2022Retrieval Augmented Classification for Long-Tail Visual Recognitioncited by 106
  15. 2021SGTR: End-to-end Scene Graph Generation with Transformercited by 102
  16. 2021The Majority Can Help The Minority: Context-rich Minority Oversampling for Long-tailed Classificationcited by 100
  17. 2022Inducing Neural Collapse in Imbalanced Learning: Do We Really Need a Learnable Classifier at the End of Deep Neural Network?cited by 100
  18. 2021MetaSAug: Meta Semantic Augmentation for Long-Tailed Visual Recognitioncited by 99
  19. 2021ResLT: Residual Learning for Long-tailed Recognitioncited by 96
  20. 2021Self-Supervised Aggregation of Diverse Experts for Test-Agnostic Long-Tailed Recognitioncited by 95
  21. 2022Balanced MSE for Imbalanced Visual Regressioncited by 89
  22. 2021ACE: Ally Complementary Experts for Solving Long-Tailed Recognition in One-Shotcited by 86
  23. 2022Temporal Action Segmentation: An Analysis of Modern Techniquescited by 85
  24. 2021Distributional Robustness Loss for Long-tail Learningcited by 84
  25. 2022Few-Shot Class-Incremental Learning from an Open-Set Perspectivecited by 83
  26. 2022Imbalance Trouble: Revisiting Neural-Collapse Geometrycited by 80
  27. 2021Detecting Human-Object Interaction via Fabricated Compositional Learningcited by 79
  28. 2021Sample Selection with Uncertainty of Losses for Learning with Noisy Labelscited by 79
  29. 2023Generating images of rare concepts using pre-trained diffusion modelscited by 78
  30. 2022ActFormer: A GAN-based Transformer towards General Action-Conditioned 3D Human Motion Generationcited by 77
  31. 2021FASA: Feature Augmentation and Sampling Adaptation for Long-Tailed Instance Segmentationcited by 76
  32. 2022A Survey on Long-Tailed Visual Recognitioncited by 75
  33. 2023Long-tailed Visual Recognition via Gaussian Clouded Logit Adjustmentcited by 75
  34. 2021Label-Imbalanced and Group-Sensitive Classification under Overparameterizationcited by 74
  35. 2021Self Supervision to Distillation for Long-Tailed Visual Recognitioncited by 73
  36. 2021Distilling Virtual Examples for Long-tailed Recognitioncited by 72
  37. 2023DELTA: degradation-free fully test-time adaptationcited by 72
  38. 2021Learning Fast Sample Re-weighting Without Reward Datacited by 69

50 most cited of 806 All 806 (JSON)