mlsanityData to 14 Oct 2026

Margin Calibration for Long-Tailed Visual Recognition

arXiv 2112.07225 ↗First posted 14 Dec 2021cs.CV cs.AI cs.LGcited by 17

Not on the map. Why?

Abstract (arXiv)

The long-tailed class distribution in visual recognition tasks poses great challenges for neural networks on how to handle the biased predictions between head and tail classes, i.e., the model tends to classify tail classes as head classes. While existing research focused on data resampling and loss function engineering, in this paper, we take a different perspective: the classification margins. We study the relationship between the margins and logits (classification scores) and empirically observe the biased margins and the biased logits are positively correlated. We propose MARC, a simple yet effective MARgin Calibration function to dynamically calibrate the biased margins for unbiased logits. We validate MARC through extensive experiments on common long-tailed benchmarks including CIFAR-LT, ImageNet-LT, Places-LT, and iNaturalist-LT. Experimental results demonstrate that our MARC achieves favorable results on these benchmarks. In addition, MARC is extremely easy to implement with just three lines of code. We hope this simple method will motivate people to rethink the biased margins and biased logits in long-tailed visual recognition.

Key facts. Cited by 17 papers in the map (data through 14 Oct 2026). Builds on 29 of its 52 references that are in the map. First posted on arXiv 14 Dec 2021.

Builds on

29 of 52 references.

References · 29

  1. 2015Deep Residual Learning for Image Recognitioncited by 52,096
  2. 2019PyTorch: An Imperative Style, High-Performance Deep Learning Librarycited by 20,305
  3. 2014Very Deep Convolutional Networks for Large-Scale Image Recognitioncited by 15,364
  4. 2017On Calibration of Modern Neural Networkscited by 4,960
  5. 2019Class-Balanced Loss Based on Effective Number of Samplescited by 1,245
  6. 2019Learning Imbalanced Datasets with Label-Distribution-Aware Margin Losscited by 989
  7. 2019Decoupling Representation and Classifier for Long-Tailed Recognitioncited by 809
  8. 2019Large-Scale Long-Tailed Recognition in an Open Worldcited by 705
  9. 2020Long-tail learning via logit adjustmentcited by 618
  10. 2017A systematic study of the class imbalance problem in convolutional neural networkscited by 513
  11. 2019BBN: Bilateral-Branch Network with Cumulative Learning for Long-Tailed Visual Recognitioncited by 403
  12. 2020Balanced Meta-Softmax for Long-Tailed Visual Recognitioncited by 400
17 more
  1. 2020Equalization Loss for Long-Tailed Object Recognitioncited by 285
  2. 2020Long-Tailed Classification by Keeping the Good and Removing the Bad Momentum Causal Effectcited by 256
  3. 2018What is the Effect of Importance Weighting in Deep Learning?cited by 251
  4. 2020Rethinking the Value of Labels for Improving Class-Imbalanced Learningcited by 216
  5. 2021Distribution Alignment: A Unified Framework for Long-tail Visual Recognitioncited by 198
  6. 2020Disentangling Label Distribution for Long-tailed Visual Recognitioncited by 172
  7. 2015Cost Sensitive Learning of Deep Feature Representations from Imbalanced Datacited by 167
  8. 2020Seesaw Loss for Long-Tailed Instance Segmentationcited by 146
  9. 2020Rethinking Class-Balanced Methods for Long-Tailed Visual Recognition from a Domain Adaptation Perspectivecited by 143
  10. 2021Contrastive Learning based Hybrid Networks for Long-Tailed Image Classificationcited by 130
  11. 2015Relay Backpropagation for Effective Learning of Deep Convolutional Neural Networkscited by 115
  12. 2018Large Margin Deep Networks for Classificationcited by 112
  13. 2017The Devil is in the Tails: Fine-grained Classification in the Wildcited by 89
  14. 2021RSG: A Simple but Effective Module for Learning Imbalanced Datasetscited by 61
  15. 2017Deep Over-sampling Framework for Classifying Imbalanced Datacited by 59
  16. 2019Adjusting Decision Boundary for Class Imbalanced Learningcited by 52
  17. 2020The Devil is the Classifier: Investigating Long Tail Relation Classification with Decoupling Analysiscited by 10

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2022: 3; 2023: 5; 2024: 6; 2025: 1; 2026: 2.

17 papers.

Sort:

By area:

  1. 2023PandaLM: An Automatic Evaluation Benchmark for LLM Instruction Tuning Optimizationcited by 316
  2. 2022USB: A Unified Semi-supervised Learning Benchmark for Classificationcited by 95
  3. 2022A Survey on Long-Tailed Visual Recognitioncited by 75
  4. 2024A Hard-to-Beat Baseline for Training-free CLIP-based Adaptationcited by 53
  5. 2023Exploring Vision-Language Models for Imbalanced Learningcited by 35
  6. 2023Understanding and Mitigating the Label Noise in Pre-training on Downstream Taskscited by 27
  7. 2022An Embarrassingly Simple Baseline for Imbalanced Semi-Supervised Learningcited by 12
  8. 2024Difficulty-aware Balancing Margin Loss for Long-tailed Recognitioncited by 9
  9. 2024Impact of Noisy Supervision in Foundation Model Learningcited by 8
  10. 2023RAHNet: Retrieval Augmented Hybrid Network for Long-tailed Graph Classificationcited by 7
  11. 2023CDMPP: A Device-Model Agnostic Framework for Latency Prediction of Tensor Programscited by 3
  12. 2024Prior2Posterior: Model Prior Correction for Long-Tailed Learningcited by 2
5 more
  1. 2024Text-Guided Mixup Towards Long-Tailed Image Categorizationcited by 1
  2. 2024CALA: A Class-Aware Logit Adapter for Few-Shot Class-Incremental Learningcited by 1
  3. 2025Mixture of Balanced Information Bottlenecks for Long-Tailed Visual Recognitioncited by 1
  4. 2026Why Not Hyperparameter-Friendly Optimisation? A Monotonic Adaptive Norm Rescaling Approach For Long-Tailed Recognitioncited by 0
  5. 2026StrokeTimer: Robust Representation Learning for Ischemic Stroke Onset-Time Estimation from Non-contrast CTcited by 0