mlsanityData to 7 Oct 2026

SASSL: Enhancing Self-Supervised Learning via Neural Style Transfer

arXiv 2312.01187 ↗First posted 2 Dec 2023cs.CV cs.LG stat.MLcited by 2

Not on the map. Why?

Abstract (arXiv)

Existing data augmentation in self-supervised learning, while diverse, fails to preserve the inherent structure of natural images. This results in distorted augmented samples with compromised semantic information, ultimately impacting downstream performance. To overcome this limitation, we propose SASSL: Style Augmentations for Self Supervised Learning, a novel data augmentation technique based on Neural Style Transfer. SASSL decouples semantic and stylistic attributes in images and applies transformations exclusively to their style while preserving content, generating diverse samples that better retain semantic information. SASSL boosts top-1 image classification accuracy on ImageNet by up to 2 percentage points compared to established self-supervised methods like MoCo, SimCLR, and BYOL, while achieving superior transfer learning performance across various datasets. Because SASSL can be performed asynchronously as part of the data augmentation pipeline, these performance impacts can be obtained with no change in pretraining throughput.

Key facts. Cited by 2 papers in the map (data through 7 Oct 2026). Builds on 48 of its 75 references that are in the map. First posted on arXiv 2 Dec 2023.

Builds on

48 of 75 references.

References · 48

  1. 2015Deep Residual Learning for Image Recognitioncited by 51,809
  2. 2018BERT: Pre-training of Deep Bidirectional Transformers for Language Understandingcited by 40,204
  3. 2014Very Deep Convolutional Networks for Large-Scale Image Recognitioncited by 15,267
  4. 2020A Simple Framework for Contrastive Learning of Visual Representationscited by 11,979
  5. 2018The Unreasonable Effectiveness of Deep Features as a Perceptual Metriccited by 11,029
  6. 2019Momentum Contrast for Unsupervised Visual Representation Learningcited by 7,482
  7. 2021Masked Autoencoders Are Scalable Vision Learnerscited by 7,339
  8. 2021Emerging Properties in Self-Supervised Vision Transformerscited by 6,893
  9. 2015Rethinking the Inception Architecture for Computer Visioncited by 5,451
  10. 2017mixup: Beyond Empirical Risk Minimizationcited by 5,155
  11. 2020Bootstrap your own latent: A new approach to self-supervised Learningcited by 4,712
  12. 2019CutMix: Regularization Strategy to Train Strong Classifiers with Localizable Featurescited by 2,875
36 more
  1. 2020Exploring Simple Siamese Representation Learningcited by 2,853
  2. 2020Unsupervised Learning of Visual Features by Contrasting Cluster Assignmentscited by 2,845
  3. 2016Perceptual Losses for Real-Time Style Transfer and Super-Resolutioncited by 2,678
  4. 2020Improved Baselines with Momentum Contrastive Learningcited by 2,282
  5. 2013Describing Textures in the Wildcited by 2,110
  6. 2017Arbitrary Style Transfer in Real-time with Adaptive Instance Normalizationcited by 1,998
  7. 2021Barlow Twins: Self-Supervised Learning via Redundancy Reductioncited by 1,818
  8. 2021An Empirical Study of Training Self-Supervised Vision Transformerscited by 1,610
  9. 2018ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustnesscited by 1,430
  10. 2017Random Erasing Data Augmentationcited by 1,235
  11. 2021VICReg: Variance-Invariance-Covariance Regularization for Self-Supervised Learningcited by 1,013
  12. 2020What Makes for Good Views for Contrastive Learning?cited by 843
  13. 2018Do Better ImageNet Models Transfer Better?cited by 560
  14. 2017Large Batch Training of Convolutional Networkscited by 529
  15. 2015A Neural Algorithm of Artistic Stylecited by 480
  16. 2017Universal Style Transfer via Feature Transformscited by 352
  17. 2016A Learned Representation For Artistic Stylecited by 340
  18. 2022Masked Siamese Networks for Label-Efficient Learningcited by 301
  19. 2017Neural Style Transfer: A Reviewcited by 199
  20. 2021StyTr2: Image Style Transfer with Transformerscited by 186
  21. 2021AdaAttN: Revisit Attention Mechanism in Arbitrary Neural Style Transfercited by 178
  22. 2019Photorealistic Style Transfer via Wavelet Transformscited by 166
  23. 2017Demystifying Neural Style Transfercited by 152
  24. 2020Demystifying Contrastive Self-Supervised Learning: Invariances, Augmentations and Dataset Biasescited by 143
  25. 2018Avatar-Net: Multi-scale Zero-shot Style Transfer by Feature Decorationcited by 115
  26. 2016Controlling Perceptual Factors in Neural Style Transfercited by 99
  27. 2015Optimal whitening and decorrelationcited by 93
  28. 2019Dynamic Instance Normalization for Arbitrary Style Transfercited by 83
  29. 2018Style Augmentation: Data Augmentation via Style Randomizationcited by 79
  30. 2017Stable and Controllable Neural Texture Synthesis and Style Transfer Using Histogram Lossescited by 69
  31. 2021Improving Transferability of Representations via Augmentation-Aware Self-Supervisioncited by 48
  32. 2019Diversified Arbitrary Style Transfer via Deep Feature Perturbationcited by 42
  33. 2020SelfAugment: Automatic Augmentation Policies for Self-Supervised Learningcited by 32
  34. 2022You Only Cut Once: Boosting Data Augmentation with a Single Cutcited by 22
  35. 2019STaDA: Style Transfer as Data Augmentationcited by 21
  36. 2022On the Importance of Hyperparameters and Data Augmentation for Self-Supervised Learningcited by 16

Built on by

Cited by year

Citing papers per year: 2024: 1; 2025: 1.

2 papers.

  1. 2024Can Generative Models Improve Self-Supervised Representation Learning?cited by 3
  2. 2025AdvDINO: Domain-Adversarial Self-Supervised Representation Learning for Spatial Proteomicscited by 1