mlsanityData to 7 Oct 2026

Discover the Unknown Biased Attribute of an Image Classifier

arXiv 2104.14556 ↗First posted 29 Apr 2021cs.CVcited by 46

Not on the map. Why?

Abstract (arXiv)

Recent works find that AI algorithms learn biases from data. Therefore, it is urgent and vital to identify biases in AI algorithms. However, the previous bias identification pipeline overly relies on human experts to conjecture potential biases (e.g., gender), which may neglect other underlying biases not realized by humans. To help human experts better find the AI algorithms' biases, we study a new problem in this work -- for a classifier that predicts a target attribute of the input image, discover its unknown biased attribute. To solve this challenging problem, we use a hyperplane in the generative model's latent space to represent an image attribute; thus, the original problem is transformed to optimizing the hyperplane's normal vector and offset. We propose a novel total-variation loss within this framework as the objective function and a new orthogonalization penalty as a constraint. The latter prevents trivial solutions in which the discovered biased attribute is identical with the target or one of the known-biased attributes. Extensive experiments on both disentanglement datasets and real-world datasets show that our method can discover biased attributes and achieve better disentanglement w.r.t. target attributes. Furthermore, the qualitative results show that our method can discover unnoticeable biased attributes for various object and scene classifiers, proving our method's generalizability for detecting biased attributes in diverse domains of images. The code is available at https://git.io/J3kMh.

Key facts. Cited by 46 papers in the map (data through 7 Oct 2026). Builds on 39 of its 54 references that are in the map. First posted on arXiv 29 Apr 2021.

Builds on

39 of 54 references.

References · 39

  1. 2015Deep Residual Learning for Image Recognitioncited by 51,809
  2. 2014Adam: A Method for Stochastic Optimizationcited by 36,960
  3. 2019PyTorch: An Imperative Style, High-Performance Deep Learning Librarycited by 20,160
  4. 2014Microsoft COCO: Common Objects in Contextcited by 17,498
  5. 2014Explaining and Harnessing Adversarial Examplescited by 7,287
  6. 2017Towards Deep Learning Models Resistant to Adversarial Attackscited by 6,326
  7. 2018A Style-Based Generator Architecture for Generative Adversarial Networkscited by 5,986
  8. 2014Deep Learning Face Attributes in the Wildcited by 3,679
  9. 2019Analyzing and Improving the Image Quality of StyleGANcited by 3,580
  10. 2017Progressive Growing of GANs for Improved Quality, Stability, and Variationcited by 2,967
  11. 2015VQA: Visual Question Answeringcited by 2,489
  12. 2016Equality of Opportunity in Supervised Learningcited by 2,000
27 more
  1. 2015LSUN: Construction of a Large-scale Image Dataset using Deep Learning with Humans in the Loopcited by 1,350
  2. 2018Model Cards for Model Reportingcited by 1,068
  3. 2018Challenging Common Assumptions in the Unsupervised Learning of Disentangled Representationscited by 929
  4. 2017Counterfactual Fairnesscited by 802
  5. 2019Interpreting the Latent Space of GANs for Semantic Face Editingcited by 691
  6. 2018Disentangling by Factorisingcited by 614
  7. 2020InterFaceGAN: Interpreting the Disentangled Face Representation Learned by GANscited by 393
  8. 2020Unsupervised Discovery of Interpretable Directions in the GAN Latent Spacecited by 288
  9. 2019Towards Fairness in Visual Recognition: Effective Strategies for Bias Mitigationcited by 252
  10. 2017Variational Inference of Disentangled Latent Concepts from Unlabeled Observationscited by 240
  11. 2019Flexibly Fair Representation Learning by Disentanglementcited by 196
  12. 2019On the Fairness of Disentangled Representationscited by 144
  13. 2019Why Can't I Dance in the Mall? Learning to Mitigate Scene Bias in Action Recognitioncited by 117
  14. 2019Fair Generative Modeling via Weak Supervisioncited by 114
  15. 2020The Hessian Penalty: A Weak Prior for Unsupervised Disentanglementcited by 90
  16. 2018Learning Disentangled Joint Continuous and Discrete Representationscited by 83
  17. 2020Fairness by Learning Orthogonal Disentangled Representationscited by 76
  18. 2019Image Counterfactual Sensitivity Analysis for Detecting Unintended Biascited by 74
  19. 2020Towards causal benchmarking of bias in face analysis algorithmscited by 64
  20. 2020Lifespan Age Transformation Synthesiscited by 60
  21. 2018Explicit Bias Discovery in Visual Question Answering Modelscited by 42
  22. 2020Gender Slopes: Counterfactual Fairness for Computer Vision Models by Attribute Manipulationcited by 36
  23. 2017Empirically Analyzing the Effect of Dataset Biases on Deep Face Recognition Systemscited by 28
  24. 2018Understanding Unequal Gender Classification Accuracy from Face Imagescited by 25
  25. 2020NestedVAE: Isolating Common Factors via Weak Supervisioncited by 13
  26. 2019Variational Learning with Disentanglement-PyTorchcited by 5
  27. 2020Assembling Semantically-Disentangled Representations for Predictive-Generative Models via Adaptation from Synthetic Domaincited by 1

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2021: 5; 2022: 17; 2023: 10; 2024: 9; 2025: 4; 2026: 1.

46 papers.

Sort:

  1. 2022A Whac-A-Mole Dilemma: Shortcuts Come in Multiples Where Mitigating One Amplifies Otherscited by 85
  2. 2022Distilling Model Failures as Directions in Latent Spacecited by 82
  3. 2023Discover and Cure: Concept-aware Mitigation of Spurious Correlationcited by 71
  4. 2023Discovering and Mitigating Visual Biases through Keyword Explanationcited by 58
  5. 2022Discover and Mitigate Unknown Biases with Debiasing Alternate Networkscited by 47
  6. 2023LANCE: Stress-testing Visual Models by Generating Language-guided Counterfactual Imagescited by 46
  7. 2021STEEX: Steering Counterfactual Explanations with Semanticscited by 40
  8. 2022Invariant Feature Learning for Generalized Long-Tailed Classificationcited by 34
  9. 2022StyleT2I: Toward Compositional and High-Fidelity Text-to-Image Synthesiscited by 33
  10. 2022On Learning Fairness and Accuracy on Multiple Subgroupscited by 31
  11. 2023Can You Rely on Your Model Evaluation? Improving Model Evaluation with Synthetic Test Datacited by 31
  12. 2022Evidence for Hypodescent in Visual Semantic AIcited by 30
34 more
  1. 2022OCTET: Object-aware Counterfactual Explanationscited by 28
  2. 2023Zero-shot Model Diagnosiscited by 23
  3. 2022Mitigating Dataset Bias by Using Per-sample Gradientcited by 21
  4. 2022On Modality Bias Recognition and Reductioncited by 20
  5. 2021Latent Space Smoothing for Individually Fair Representationscited by 18
  6. 2021Anatomizing Bias in Facial Analysiscited by 18
  7. 2023Identification of Systematic Errors of Image Classifiers on Rare Subgroupscited by 17
  8. 2022Learning to Split for Automatic Bias Detectioncited by 16
  9. 2024Fairness and Bias Mitigation in Computer Vision: A Surveycited by 12
  10. 2024Discover and Mitigate Multiple Biased Subgroups in Image Classifierscited by 11
  11. 2022medXGAN: Visual Explanations for Medical Classifiers through a Generative Latent Spacecited by 10
  12. 2022Chairs Can be Stood on: Overcoming Object Bias in Human-Object Interaction Detectioncited by 9
  13. 2022Self-supervised debiasing using low rank regularizationcited by 8
  14. 2021Learning Stable Classifiers by Transferring Unstable Featurescited by 6
  15. 2022Towards a More Rigorous Science of Blindspot Discovery in Image Classification Modelscited by 5
  16. 2024Intersectional Unfairness Discoverycited by 5
  17. 2022Explaining Image Classifiers Using Contrastive Counterfactuals in Generative Latent Spacescited by 4
  18. 2023Improving Out-of-Distribution Robustness of Classifiers via Generative Interpolationcited by 4
  19. 2023Identifying Systematic Errors in Object Detectors with the SCROD Pipelinecited by 4
  20. 2024Unsupervised Model Diagnosiscited by 4
  21. 2025Classifier-to-Bias: Toward Unsupervised Automatic Bias Detection for Visual Classifierscited by 4
  22. 2022A survey of Identification and mitigation of Machine Learning algorithmic biases in Image Analysiscited by 3
  23. 2023Cross-Modal Attribute Insertions for Assessing the Robustness of Vision-and-Language Learningcited by 3
  24. 2025Behind the Screens: Uncovering Bias in AI-Driven Video Interview Assessments Using Counterfactualscited by 2
  25. 2021Explainability Requires Interactivitycited by 1
  26. 2023Revealing Model Biases: Assessing Deep Neural Networks via Recovered Sample Analysiscited by 1
  27. 2024Common-Sense Bias Modeling for Classification Taskscited by 1
  28. 2024Identification of Fine-grained Systematic Errors via Controlled Scene Generationcited by 0
  29. 2024Towards Real-world Debiasing: Rethinking Evaluation, Challenge, and Solutioncited by 0
  30. 2024Top-GAP: Integrating Size Priors in CNNs for more Interpretability, Robustness, and Bias Mitigationcited by 0
  31. 2024Detecting Unforeseen Data Properties with Diffusion Autoencoder Embeddings using Spine MRI datacited by 0
  32. 2025A Critical Review of Predominant Bias in Neural Networkscited by 0
  33. 2025Multi Attribute Bias Mitigation via Representation Learningcited by 0
  34. 2026Mitigating Bias in Concept Bottleneck Models for Fair and Interpretable Image Classificationcited by 0