mlsanityData to 14 Oct 2026

Correlated Input-Dependent Label Noise in Large-Scale Image Classification

arXiv 2105.10305 ↗First posted 19 May 2021cs.LG cs.CV stat.MLcited by 37

Not on the map. Why?

Abstract (arXiv)

Large scale image classification datasets often contain noisy labels. We take a principled probabilistic approach to modelling input-dependent, also known as heteroscedastic, label noise in these datasets. We place a multivariate Normal distributed latent variable on the final hidden layer of a neural network classifier. The covariance matrix of this latent variable, models the aleatoric uncertainty due to label noise. We demonstrate that the learned covariance structure captures known sources of label noise between semantically similar and co-occurring classes. Compared to standard neural network training and other baselines, we show significantly improved accuracy on Imagenet ILSVRC 2012 79.3% (+2.6%), Imagenet-21k 47.0% (+1.1%) and JFT 64.7% (+1.6%). We set a new state-of-the-art result on WebVision 1.0 with 76.6% top-1 accuracy. These datasets range from over 1M to over 300M training examples and from 1k classes to more than 21k classes. Our method is simple to use, and we provide an implementation that is a drop-in replacement for the final fully-connected layer in a deep classifier.

Key facts. Cited by 37 papers in the map (data through 14 Oct 2026). Builds on 35 of its 49 references that are in the map. First posted on arXiv 19 May 2021.

Builds on

35 of 49 references.

References · 35

  1. 2015Deep Residual Learning for Image Recognitioncited by 52,096
  2. 2015Distilling the Knowledge in a Neural Networkcited by 10,324
  3. 2019EfficientNet: Rethinking Model Scaling for Convolutional Neural Networkscited by 6,442
  4. 2017mixup: Beyond Empirical Risk Minimizationcited by 5,183
  5. 2017On Calibration of Modern Neural Networkscited by 4,960
  6. 2015Dropout as a Bayesian Approximation: Representing Model Uncertainty in Deep Learningcited by 4,448
  7. 2016Simple and Scalable Predictive Uncertainty Estimation using Deep Ensemblescited by 3,926
  8. 2016Categorical Reparameterization with Gumbel-Softmaxcited by 2,739
  9. 2017What Uncertainties Do We Need in Bayesian Deep Learning for Computer Vision?cited by 2,352
  10. 2016Xception: Deep Learning with Depthwise Separable Convolutionscited by 2,332
  11. 2016Inception-v4, Inception-ResNet and the Impact of Residual Connections on Learningcited by 1,512
  12. 2019Can You Trust Your Model's Uncertainty? Evaluating Predictive Uncertainty Under Dataset Shiftcited by 1,287
23 more
  1. 2016The Concrete Distribution: A Continuous Relaxation of Discrete Random Variablescited by 1,148
  2. 2018Co-teaching: Robust Training of Deep Neural Networks with Extremely Noisy Labelscited by 1,059
  3. 2017A Closer Look at Memorization in Deep Networkscited by 942
  4. 2017Revisiting Unreasonable Effectiveness of Data in Deep Learning Eracited by 919
  5. 2019Big Transfer (BiT): General Visual Representation Learningcited by 769
  6. 2017MentorNet: Learning Data-Driven Curriculum for Very Deep Neural Networks on Corrupted Labelscited by 653
  7. 2015Weight Uncertainty in Neural Networkscited by 554
  8. 2020Bayesian Deep Learning and a Probabilistic Perspective of Generalizationcited by 480
  9. 2019A Large-scale Study of Representation Learning with the Visual Task Adaptation Benchmarkcited by 425
  10. 2019Transfusion: Understanding Transfer Learning for Medical Imagingcited by 368
  11. 2020Are we done with ImageNet?cited by 333
  12. 2014Training Deep Neural Networks on Noisy Labels with Bootstrappingcited by 322
  13. 2017WebVision Database: Visual Learning and Understanding from Web Datacited by 262
  14. 2017Concrete Dropoutcited by 227
  15. 2019Evaluating Scalable Bayesian Deep Learning Methods for Robust Computer Visioncited by 182
  16. 2017CleanNet: Transfer Learning for Scalable Image Classifier Training with Label Noisecited by 181
  17. 2020Efficient and Scalable Bayesian Neural Nets with Rank-1 Factorscited by 171
  18. 2018CurriculumNet: Weakly Supervised Learning from Large-Scale Web Imagescited by 126
  19. 2020Stochastic Segmentation Networks: Modelling Spatially Correlated Aleatoric Uncertaintycited by 101
  20. 2020On Robustness and Transferability of Convolutional Neural Networkscited by 98
  21. 2020Heteroskedastic and Imbalanced Deep Learning with Adaptive Regularizationcited by 49
  22. 2020Combining Ensembles and Data Augmentation can Harm your Calibrationcited by 44
  23. 2014Mind the Nuisance: Gaussian Process Classification using Privileged Noisecited by 7

Built on by

Cited by year (2026 to 14 Oct)

Citing papers per year: 2021: 7; 2022: 7; 2023: 12; 2024: 4; 2025: 4; 2026: 3.

37 papers.

Sort:

By area:

  1. 2023Scaling Vision Transformers to 22 Billion Parameterscited by 694
  2. 2021Epistemic Neural Networkscited by 120
  3. 2022Plex: Towards Reliability using Pretrained Large Model Extensionscited by 102
  4. 2021Open-Vocabulary Instance Segmentation via Robust Cross-Modal Pseudo-Labelingcited by 84
  5. 2021Uncertainty Baselines: Benchmarks for Uncertainty & Robustness in Deep Learningcited by 82
  6. 2024Variational Bayesian Last Layerscited by 69
  7. 2023Uncertainty Estimation by Fisher Information-based Evidential Deep Learningcited by 58
  8. 2022Learning with Neighbor Consistency for Noisy Labelscited by 55
  9. 2022Is one annotation enough? A data-centric image classification benchmark for noisy and ambiguous label estimationcited by 43
  10. 2022CMW-Net: Learning a Class-Aware Sample Weighting Mapping for Robust Deep Learningcited by 24
  11. 2023Improving Image Recognition by Retrieving from Web-Scale Image-Text Datacited by 20
  12. 2022Transfer and Marginalize: Explaining Away Label Noise with Privileged Informationcited by 15
25 more
  1. 2021Constrained Instance and Class Reweighting for Robust Learning under Label Noisecited by 12
  2. 2023Density-Softmax: Efficient Test-time Model for Uncertainty Estimation and Robustness under Distribution Shiftscited by 12
  3. 2023Massively Scaling Heteroscedastic Classifierscited by 11
  4. 2023When does Privileged Information Explain Away Label Noise?cited by 10
  5. 2021Deep Classifiers with Label Noise Modeling and Distance Awarenesscited by 8
  6. 2024Density-Regression: Efficient and Distance-Aware Deep Regressor for Uncertainty Estimation under Distribution Shiftscited by 7
  7. 2024Selectively Answering Visual Questionscited by 5
  8. 2024Knockout: A simple way to handle missing inputscited by 4
  9. 2025Benchmarking Uncertainty and its Disentanglement in multi-label Chest X-Ray Classificationcited by 4
  10. 2022Cold Posteriors through PAC-Bayescited by 3
  11. 2023Logistic-Normal Likelihoods for Heteroscedastic Label Noisecited by 3
  12. 2023Label Smarter, Not Harder: CleverLabel for Faster Annotation of Ambiguous Image Classification with Higher Qualitycited by 3
  13. 2025Uncertainty-Aware Trajectory Prediction via Rule-Regularized Heteroscedastic Deep Classificationcited by 3
  14. 2022Tackling Instance-Dependent Label Noise with Dynamic Distribution Calibrationcited by 2
  15. 2023Pushing the Accuracy-Group Robustness Frontier with Introspective Self-playcited by 2
  16. 2023Pi-DUAL: Using Privileged Information to Distinguish Clean from Noisy Labelscited by 2
  17. 2023Transferable Candidate Proposal with Bounded Uncertaintycited by 2
  18. 2025Rethinking Approximate Gaussian Inference in Classificationcited by 2
  19. 2026Can LLMs Learn to Reason Robustly under Noisy Supervision?cited by 2
  20. 2025On the Generalization of Representation Uncertainty in Earth Observationcited by 1
  21. 2021Life is not black and white -- Combining Semi-Supervised Learning with fuzzy labelscited by 0
  22. 2021Dense Uncertainty Estimation via an Ensemble-based Conditional Latent Variable Modelcited by 0
  23. 2023FPR Estimation for Fraud Detection in the Presence of Class-Conditional Label Noisecited by 0
  24. 2026Fisher8: Stabilizing Neural Heteroscedastic Regression via Output-Layer Fisher Geometrycited by 0
  25. 2026Flexible Spectral-Normalized Neural Gaussian Process for Dynamic Aperture Predictioncited by 0