On the Fly Neural Style Smoothing for Risk-Averse Domain Generalization
Not on the map. Why?
Abstract (arXiv)
Achieving high accuracy on data from domains unseen during training is a fundamental challenge in domain generalization (DG). While state-of-the-art DG classifiers have demonstrated impressive performance across various tasks, they have shown a bias towards domain-dependent information, such as image styles, rather than domain-invariant information, such as image content. This bias renders them unreliable for deployment in risk-sensitive scenarios such as autonomous driving where a misclassification could lead to catastrophic consequences. To enable risk-averse predictions from a DG classifier, we propose a novel inference procedure, Test-Time Neural Style Smoothing (TT-NSS), that uses a "style-smoothed" version of the DG classifier for prediction at test time. Specifically, the style-smoothed classifier classifies a test image as the most probable class predicted by the DG classifier on random re-stylizations of the test image. TT-NSS uses a neural style transfer module to stylize a test image on the fly, requires only black-box access to the DG classifier, and crucially, abstains when predictions of the DG classifier on the stylized test images lack consensus. Additionally, we propose a neural style smoothing (NSS) based training procedure that can be seamlessly integrated with existing DG methods. This procedure enhances prediction consistency, improving the performance of TT-NSS on non-abstained samples. Our empirical results demonstrate the effectiveness of TT-NSS and NSS at producing and improving risk-averse predictions on unseen domains from DG classifiers trained with SOTA training methods on various benchmark datasets and their variations.
Key facts. Cited by 3 papers in the map (data through 14 Oct 2026). Builds on 63 of its 92 references that are in the map. First posted on arXiv 17 Jul 2023.
Builds on
63 of 92 references.
References · 63
- 2014Microsoft COCO: Common Objects in Contextcited by 17,608
- 2014Very Deep Convolutional Networks for Large-Scale Image Recognitioncited by 15,364
- 2017mixup: Beyond Empirical Risk Minimizationcited by 5,183
- 2013Intriguing properties of neural networkscited by 5,155
- 2015Domain-Adversarial Training of Neural Networkscited by 3,044
- 2019Benchmarking Neural Network Robustness to Common Corruptions and Perturbationscited by 2,877
- 2016Perceptual Losses for Real-Time Style Transfer and Super-Resolutioncited by 2,694
- 2016A Baseline for Detecting Misclassified and Out-of-Distribution Examples in Neural Networkscited by 2,385
- 2020FixMatch: Simplifying Semi-Supervised Learning with Consistency and Confidencecited by 2,085
- 2017Arbitrary Style Transfer in Real-time with Adaptive Instance Normalizationcited by 2,010
- 2019Invariant Risk Minimizationcited by 1,854
- 2018ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustnesscited by 1,432
38 more
- 2019Certified Adversarial Robustness via Randomized Smoothingcited by 1,340
- 2018Obfuscated Gradients Give a False Sense of Security: Circumventing Defenses to Adversarial Examplescited by 1,268
- 2017Deep Hashing Network for Unsupervised Domain Adaptationcited by 1,188
- 2020In Search of Lost Domain Generalizationcited by 1,054
- 2017Deeper, Broader and Artier Domain Generalizationcited by 986
- 2018Deep Anomaly Detection with Outlier Exposurecited by 942
- 2019AugMix: A Simple Data Processing Method to Improve Robustness and Uncertaintycited by 901
- 2021Generalizing to Unseen Domains: A Survey on Domain Generalizationcited by 766
- 2017ZOO: Zeroth Order Optimization based Black-box Attacks to Deep Neural Networks without Training Substitute Modelscited by 666
- 2018Recognition in Terra Incognitacited by 658
- 2021Domain Generalization with MixStylecited by 566
- 2018Black-box Adversarial Attacks with Limited Queries and Informationcited by 501
- 2018Certified Robustness to Adversarial Examples with Differential Privacycited by 468
- 2018Efficient Neural Network Robustness Certification with General Activation Functionscited by 445
- 2020Self-Challenging Improves Cross-Domain Generalizationcited by 426
- 2021MEMO: Test Time Robustness via Adaptation and Augmentationcited by 377
- 2016A Learned Representation For Artistic Stylecited by 341
- 2018Why ReLU networks yield high-confidence predictions far away from the training data and how to mitigate the problemcited by 323
- 2018Certified Defenses against Adversarial Examplescited by 306
- 2020Learning to Learn Single Domain Generalizationcited by 296
- 2018Generating Adversarial Examples with Adversarial Networkscited by 258
- 2017Improved Texture Networks: Maximizing Quality and Diversity in Feed-forward Stylization and Texture Synthesiscited by 215
- 2016Texture Networks: Feed-forward Synthesis of Textures and Stylized Imagescited by 174
- 2017EAD: Elastic-Net Attacks to Deep Neural Networks via Adversarial Examplescited by 155
- 2018Strike (with) a Pose: Neural Networks Are Easily Fooled by Strange Poses of Familiar Objectscited by 143
- 2019Generalizing to unseen domains via distribution matchingcited by 110
- 2016Controlling Perceptual Factors in Neural Style Transfercited by 100
- 2018Benchmarking Neural Network Robustness to Common Corruptions and Surface Variationscited by 99
- 2021Towards Principled Disentanglement for Domain Generalizationcited by 91
- 2021AugMax: Adversarial Composition of Random Augmentations for Robust Trainingcited by 90
- 2022Style-Hallucinated Dual Consistency Learning for Domain Generalized Semantic Segmentationcited by 89
- 2019Support and Invertibility in Domain-Invariant Representationscited by 87
- 2020Consistency Regularization for Certified Robustness of Smoothed Classifierscited by 73
- 2021On the effectiveness of adversarial training against common corruptionscited by 71
- 2022Adversarial Style Augmentation for Domain Generalized Urban-Scene Segmentationcited by 70
- 2020On the Transfer of Disentangled Representations in Realistic Settingscited by 63
- 2020Classification with Rejection Based on Cost-sensitive Classificationcited by 55
- 2020Certified Defense to Image Transformations via Randomized Smoothingcited by 51
50 most cited of 63 All 63 (JSON)
Built on by
Citing papers per year: 2024: 3.
3 papers.