mlsanityData to 7 Oct 2026

Improving robustness against common corruptions with frequency biased models

arXiv 2103.16241 ↗First posted 30 Mar 2021cs.CVcited by 34

Not on the map. Why?

Abstract (arXiv)

CNNs perform remarkably well when the training and test distributions are i.i.d, but unseen image corruptions can cause a surprisingly large drop in performance. In various real scenarios, unexpected distortions, such as random noise, compression artefacts, or weather distortions are common phenomena. Improving performance on corrupted images must not result in degraded i.i.d performance - a challenge faced by many state-of-the-art robust approaches. Image corruption types have different characteristics in the frequency spectrum and would benefit from a targeted type of data augmentation, which, however, is often unknown during training. In this paper, we introduce a mixture of two expert models specializing in high and low-frequency robustness, respectively. Moreover, we propose a new regularization scheme that minimizes the total variation (TV) of convolution feature-maps to increase high-frequency robustness. The approach improves on corrupted images without degrading in-distribution performance. We demonstrate this on ImageNet-C and also for real-world corruptions on an automotive dataset, both for object classification and object detection.

Key facts. Cited by 34 papers in the map (data through 7 Oct 2026). Builds on 33 of its 45 references that are in the map. First posted on arXiv 30 Mar 2021.

Builds on

33 of 45 references.

References · 33

  1. 2015Deep Residual Learning for Image Recognitioncited by 51,809
  2. 2015Faster R-CNN: Towards Real-Time Object Detection with Region Proposal Networkscited by 9,617
  3. 2017Towards Deep Learning Models Resistant to Adversarial Attackscited by 6,326
  4. 2013Intriguing properties of neural networkscited by 5,114
  5. 2019Benchmarking Neural Network Robustness to Common Corruptions and Perturbationscited by 2,857
  6. 2020The Many Faces of Robustness: A Critical Analysis of Out-of-Distribution Generalizationcited by 1,842
  7. 2019Do ImageNet Classifiers Generalize to ImageNet?cited by 1,507
  8. 2017Accurate, Large Minibatch SGD: Training ImageNet in 1 Hourcited by 1,479
  9. 2018ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustnesscited by 1,430
  10. 2019MMDetection: Open MMLab Detection Toolbox and Benchmarkcited by 1,379
  11. 2015DeepFool: a simple and accurate method to fool deep neural networkscited by 1,315
  12. 2019Self-training with Noisy Student improves ImageNet classificationcited by 1,304
21 more
  1. 2018BDD100K: A Diverse Driving Dataset for Heterogeneous Multitask Learningcited by 1,282
  2. 2019AugMix: A Simple Data Processing Method to Improve Robustness and Uncertaintycited by 897
  3. 2015The Limitations of Deep Learning in Adversarial Settingscited by 784
  4. 2018AutoAugment: Learning Augmentation Policies from Datacited by 728
  5. 2020Fast is better than free: Revisiting adversarial trainingcited by 723
  6. 2017ZOO: Zeroth Order Optimization based Black-box Attacks to Deep Neural Networks without Training Substitute Modelscited by 656
  7. 2017Adversarial Examples Are Not Easily Detected: Bypassing Ten Detection Methodscited by 542
  8. 2017Decision-Based Adversarial Attacks: Reliable Attacks Against Black-Box Machine Learning Modelscited by 459
  9. 2020Measuring Robustness to Natural Distribution Shifts in Image Classificationcited by 452
  10. 2020Improving robustness against common corruptions by covariate shift adaptationcited by 427
  11. 2019Using Pre-Training Can Improve Model Robustness and Uncertaintycited by 373
  12. 2019Benchmarking Robustness in Object Detection: Autonomous Driving when Winter is Comingcited by 332
  13. 2018Feature Denoising for Improving Adversarial Robustnesscited by 312
  14. 2019A Fourier Perspective on Model Robustness in Computer Visioncited by 288
  15. 2018Generalisation in humans and deep neural networkscited by 237
  16. 2019Efficient Decision-based Black-box Adversarial Attacks on Face Recognitioncited by 173
  17. 2019Adversarial Examples Are a Natural Consequence of Test Error in Noisecited by 114
  18. 2019Benchmarking the Robustness of Semantic Segmentation Modelscited by 111
  19. 2016Examining the Impact of Blur on Recognition by Convolutional Networkscited by 58
  20. 2020DAWN: Vehicle Detection in Adverse Weather Nature Datasetcited by 41
  21. 2019Robustness properties of Facebook's ResNeXt WSL modelscited by 17

Built on by

Cited by year (2026 to 7 Oct)

Citing papers per year: 2021: 2; 2022: 5; 2023: 11; 2024: 7; 2025: 6; 2026: 3.

34 papers.

Sort:

By area:

  1. 2025DriveSuprim: Towards Precise Trajectory Selection for End-to-End Planningcited by 91
  2. 2022Assaying Out-Of-Distribution Generalization in Transfer Learningcited by 53
  3. 2021PRIME: A few primitives can boost robustness to common corruptionscited by 40
  4. 2023WEDGE: A multi-weather autonomous driving dataset built from generative vision-language modelscited by 28
  5. 2023A Survey on the Robustness of Computer Vision Models against Common Corruptionscited by 20
  6. 2024Can Biases in ImageNet Models Explain Generalization?cited by 20
  7. 2024Fourier-basis Functions to Bridge Augmentation Gap: Rethinking Frequency Augmentation in Image Classificationcited by 18
  8. 2023Classification robustness to common optical aberrationscited by 13
  9. 2022Improving Robustness by Enhancing Weak Subnetscited by 12
  10. 2022Continual BatchNorm Adaptation (CBNA) for Semantic Segmentationcited by 12
  11. 2023HybridAugment++: Unified Frequency Spectra Perturbations for Model Robustnesscited by 12
  12. 2024Benchmarking the Robustness of Temporal Action Detection Models Against Temporal Corruptionscited by 12
22 more
  1. 2022How Does Frequency Bias Affect the Robustness of Neural Image Classifiers against Common Corruption and Adversarial Perturbations?cited by 11
  2. 2024Dynamic Sparse Training versus Dense Training: The Unexpected Winner in Image Corruption Robustnesscited by 11
  3. 2023An Extended Study of Human-like Behavior under Adversarial Trainingcited by 9
  4. 2023Improving Model Generalization by On-manifold Adversarial Augmentation in the Frequency Domaincited by 5
  5. 2025Robust Asymmetric Heterogeneous Federated Learning with Corrupted Clientscited by 5
  6. 2023Assessment Framework for Deepfake Detection in Real-world Situationscited by 4
  7. 2022Addressing Mistake Severity in Neural Networks with Semantic Knowledgecited by 3
  8. 2021Estimating the Robustness of Classification Models by the Structure of the Learned Feature-Spacecited by 2
  9. 2023Fourier Analysis on Robustness of Graph Convolutional Neural Networks for Skeleton-based Action Recognitioncited by 2
  10. 2024A Modular System for Enhanced Robustness of Multimedia Understanding Networks via Deep Parametric Estimationcited by 2
  11. 2025DGSSA: Domain generalization with structural and stylistic augmentation for retinal vessel segmentationcited by 2
  12. 2025Examining the Impact of Optical Aberrations to Image Classification and Object Detection Modelscited by 2
  13. 2023Fourier Sensitivity and Regularization of Computer Vision Modelscited by 1
  14. 2024FFT-based Selection and Optimization of Statistics for Robust Recognition of Severely Corrupted Imagescited by 1
  15. 2024Adaptive Sensitivity Analysis for Robust Augmentation against Natural Corruptions in Image Segmentationcited by 1
  16. 2025A Validation Strategy for Deep Learning Models: Evaluating and Enhancing Robustnesscited by 1
  17. 2023t-RAIN: Robust generalization under weather-aliasing label shift attackscited by 0
  18. 2023Exploiting Frequency Spectrum of Adversarial Images for General Robustnesscited by 0
  19. 2025Provenance-Driven Reliable Semantic Medical Image Vector Reconstruction via Lightweight Blockchain-Verified Latent Fingerprintscited by 0
  20. 2026Boosting Robust AIGI Detection with LoRA-based Pairwise Trainingcited by 0
  21. 2026Robustness Emerges Early in Training Dynamics, but Is Not Preservedcited by 0
  22. 2026Suppress and Diversify: Refining Robust Pathways for Corruption Robustnesscited by 0