mlsanityData to 7 Oct 2026

Simple Disentanglement of Style and Content in Visual Representations

arXiv 2302.09795 ↗First posted 20 Feb 2023cs.LG cs.CV stat.MLcited by 10

Not on the map. Why?

Abstract (arXiv)

Learning visual representations with interpretable features, i.e., disentangled representations, remains a challenging problem. Existing methods demonstrate some success but are hard to apply to large-scale vision datasets like ImageNet. In this work, we propose a simple post-processing framework to disentangle content and style in learned representations from pre-trained vision models. We model the pre-trained features probabilistically as linearly entangled combinations of the latent content and style factors and develop a simple disentanglement algorithm based on the probabilistic model. We show that the method provably disentangles content and style features and verify its efficacy empirically. Our post-processed features yield significant domain generalization performance improvements when the distribution shift occurs due to style changes or style-related spurious correlations.

Key facts. Cited by 10 papers in the map (data through 7 Oct 2026). Builds on 40 of its 55 references that are in the map. First posted on arXiv 20 Feb 2023.

Builds on

40 of 55 references.

References · 40

  1. 2015Deep Residual Learning for Image Recognitioncited by 51,809
  2. 2014ImageNet Large Scale Visual Recognition Challengecited by 8,441
  3. 2018Representation Learning with Contrastive Predictive Codingcited by 7,662
  4. 2021Masked Autoencoders Are Scalable Vision Learnerscited by 7,339
  5. 2020Bootstrap your own latent: A new approach to self-supervised Learningcited by 4,712
  6. 2019Benchmarking Neural Network Robustness to Common Corruptions and Perturbationscited by 2,857
  7. 2020Exploring Simple Siamese Representation Learningcited by 2,853
  8. 2017Arbitrary Style Transfer in Real-time with Adaptive Instance Normalizationcited by 1,998
  9. 2019Invariant Risk Minimizationcited by 1,832
  10. 2018ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustnesscited by 1,430
  11. 2020WILDS: A Benchmark of in-the-Wild Distribution Shiftscited by 1,366
  12. 2017Interpretability Beyond Feature Attribution: Quantitative Testing with Concept Activation Vectors (TCAV)cited by 1,309
28 more
  1. 2019Energy and Policy Considerations for Deep Learning in NLPcited by 1,242
  2. 2019Distributionally Robust Neural Networks for Group Shifts: On the Importance of Regularization for Worst-Case Generalizationcited by 1,232
  3. 2015Unsupervised Visual Representation Learning by Context Predictioncited by 1,062
  4. 2018Challenging Common Assumptions in the Unsupervised Learning of Disentangled Representationscited by 929
  5. 2020Simple and Deep Graph Convolutional Networkscited by 857
  6. 2019EDA: Easy Data Augmentation Techniques for Boosting Performance on Text Classification Taskscited by 800
  7. 2018Recognition in Terra Incognitacited by 655
  8. 2021Self-Supervised Learning with Data Augmentations Provably Isolates Content from Stylecited by 308
  9. 2018Diverse Image-to-Image Translation via Disentangled Representationscited by 296
  10. 2018Towards a Definition of Disentangled Representationscited by 294
  11. 2016Unsupervised Feature Extraction by Time-Contrastive Learning and Nonlinear ICAcited by 270
  12. 2020Weakly-Supervised Disentanglement Without Compromisescited by 237
  13. 2021Contrastive Learning Inverts the Data Generating Processcited by 220
  14. 2015Large-scale Classification of Fine-Art Paintings: Learning The Right Metric on The Right Featurecited by 181
  15. 2019On the Fairness of Disentangled Representationscited by 144
  16. 2018Disentangled Representation Learning for Non-Parallel Text Style Transfercited by 127
  17. 2021A Survey on Data Augmentation for Text Classificationcited by 111
  18. 2019Learning Disentangled Representations for Recommendationcited by 110
  19. 2017Multi-Level Variational Autoencoder: Learning Disentangled Representations from Grouped Observationscited by 100
  20. 2019Weakly Supervised Disentanglement with Guaranteescited by 82
  21. 2018How to Read Paintings: Semantic Art Understanding with Multi-Modal Retrievalcited by 73
  22. 2021Post-processing for Individual Fairnesscited by 61
  23. 2021Self-Supervised Learning Disentangled Group Representation as Featurecited by 47
  24. 2021ALADIN: All Layer Adaptive Instance Normalization for Fine-grained Style Similaritycited by 34
  25. 2019Weakly Supervised Disentanglement by Pairwise Similaritiescited by 22
  26. 2019Disentangling Content and Style via Unsupervised Geometry Distillationcited by 9
  27. 2021Rethinking Content and Style: Exploring Bias for Unsupervised Disentanglementcited by 7
  28. 2020Adversarial Disentanglement with Grouped Observationscited by 5

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2023: 1; 2024: 2; 2025: 3; 2026: 4.

10 papers.

  1. 2024StyleShot: A Snapshot on Any Stylecited by 60
  2. 2025Towards LLMs Robustness to Changes in Prompt Format Stylescited by 21
  3. 2023Substance or Style: What Does Your Image Embedding Know?cited by 5
  4. 2026Inspiration Seeds: Learning Non-Literal Visual Combinations for Generative Explorationcited by 1
  5. 2024DualContrast: Unsupervised Disentangling of Content and Transformations with Implicit Parameterizationcited by 0
  6. 2025ConceptVAE: Self-Supervised Fine-Grained Concept Disentanglement from 2D Echocardiographiescited by 0
  7. 2025Disentanglement of Biological and Technical Factors via Latent Space Rotation in Clinical Imaging Improves Disease Pattern Discoverycited by 0
  8. 2026Mitigating Shortcut Learning via Feature Disentanglement in Medical Imaging: A Benchmark Studycited by 0
  9. 2026Measurement-Gated Provenance Attenuation for Frozen EEG Representationscited by 0
  10. 2026CLeaR: A Unified Framework for Resolving the Leakage-Degradation Dilemma in Style Transfercited by 0