mlsanityData to 7 Oct 2026

Investigating and Explaining the Frequency Bias in Image Classification

arXiv 2205.03154 ↗First posted 6 May 2022cs.CV cs.LG eess.IVcited by 8

Not on the map. Why?

Abstract (arXiv)

CNNs exhibit many behaviors different from humans, one of which is the capability of employing high-frequency components. This paper discusses the frequency bias phenomenon in image classification tasks: the high-frequency components are actually much less exploited than the low- and mid-frequency components. We first investigate the frequency bias phenomenon by presenting two observations on feature discrimination and learning priority. Furthermore, we hypothesize that (i) the spectral density, (ii) class consistency directly affect the frequency bias. Specifically, our investigations verify that the spectral density of datasets mainly affects the learning priority, while the class consistency mainly affects the feature discrimination.

Key facts. Cited by 8 papers in the map (data through 7 Oct 2026). Builds on 10 of its 16 references that are in the map. First posted on arXiv 6 May 2022.

Builds on

10 of 16 references.

References · 10

  1. 2015Deep Residual Learning for Image Recognitioncited by 51,809
  2. 2017mixup: Beyond Empirical Risk Minimizationcited by 5,155
  3. 2019CutMix: Regularization Strategy to Train Strong Classifiers with Localizable Featurescited by 2,875
  4. 2016Understanding deep learning requires rethinking generalizationcited by 1,657
  5. 2018ImageNet-trained CNNs are biased towards texture; increasing shape bias improves accuracy and robustnesscited by 1,430
  6. 2019High Frequency Component Helps Explain the Generalization of Convolutional Neural Networkscited by 310
  7. 2020Watch your Up-Convolution: CNN Based Generative Deep Neural Networks are Failing to Reproduce Spectral Distributionscited by 278
  8. 2019Fourier Spectrum Discrepancies in Deep Network Generated Imagescited by 138
  9. 2017Measuring the tendency of CNNs to Learn Surface Statistical Regularitiescited by 87
  10. 2021On the Frequency Bias of Generative Modelscited by 60

Built on by

Cited by year (2026 to 7 Oct)

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

8 papers.

  1. 2025Frequency-Aligned Knowledge Distillation for Lightweight Spatiotemporal Forecastingcited by 53
  2. 2025Noise Matters: Optimizing Matching Noise for Diffusion Classifierscited by 7
  3. 2024Exploring connections of spectral analysis and transfer learning in medical imagingcited by 1
  4. 2025A Gift from the Integration of Discriminative and Diffusion-based Generative Learning: Boundary Refinement Remote Sensing Semantic Segmentationcited by 1
  5. 2023Exploiting Frequency Spectrum of Adversarial Images for General Robustnesscited by 0
  6. 2025Interpreting Adversarial Attacks and Defences using Architectures with Enhanced Interpretabilitycited by 0
  7. 2026DuFal: Dual-Frequency-Aware Learning for High-Fidelity Extremely Sparse-view CBCT Reconstructioncited by 0
  8. 2026Low-Frequency Shortcuts in Texture-Driven Visual Learningcited by 0