mlsanityData to 14 Oct 2026

Style Spectroscope: Improve Interpretability and Controllability through Fourier Analysis

arXiv 2208.06140 ↗First posted 12 Aug 2022cs.CV cs.LG eess.IVcited by 3

Not on the map. Why?

Abstract (arXiv)

Universal style transfer (UST) infuses styles from arbitrary reference images into content images. Existing methods, while enjoying many practical successes, are unable of explaining experimental observations, including different performances of UST algorithms in preserving the spatial structure of content images. In addition, methods are limited to cumbersome global controls on stylization, so that they require additional spatial masks for desired stylization. In this work, we provide a systematic Fourier analysis on a general framework for UST. We present an equivalent form of the framework in the frequency domain. The form implies that existing algorithms treat all frequency components and pixels of feature maps equally, except for the zero-frequency component. We connect Fourier amplitude and phase with Gram matrices and a content reconstruction loss in style transfer, respectively. Based on such equivalence and connections, we can thus interpret different structure preservation behaviors between algorithms with Fourier phase. Given the interpretations we have, we propose two manipulations in practice for structure preservation and desired stylization. Both qualitative and quantitative experiments demonstrate the competitive performance of our method against the state-of-the-art methods. We also conduct experiments to demonstrate (1) the abovementioned equivalence, (2) the interpretability based on Fourier amplitude and phase and (3) the controllability associated with frequency components.

Key facts. Cited by 3 papers in the map (data through 14 Oct 2026). Builds on 22 of its 35 references that are in the map. First posted on arXiv 12 Aug 2022.

Builds on

22 of 35 references.

References · 22

  1. 2014Microsoft COCO: Common Objects in Contextcited by 17,608
  2. 2014Very Deep Convolutional Networks for Large-Scale Image Recognitioncited by 15,364
  3. 2018The Unreasonable Effectiveness of Deep Features as a Perceptual Metriccited by 11,177
  4. 2016Perceptual Losses for Real-Time Style Transfer and Super-Resolutioncited by 2,694
  5. 2017Arbitrary Style Transfer in Real-time with Adaptive Instance Normalizationcited by 2,010
  6. 2020FDA: Fourier Domain Adaptation for Semantic Segmentationcited by 541
  7. 2017Universal Style Transfer via Feature Transformscited by 354
  8. 2016A Learned Representation For Artistic Stylecited by 341
  9. 2018Arbitrary Style Transfer with Style-Attentional Networkscited by 198
  10. 2019Photorealistic Style Transfer via Wavelet Transformscited by 166
  11. 2018A Closed-form Solution to Photorealistic Image Stylizationcited by 133
  12. 2018Avatar-Net: Multi-scale Zero-shot Style Transfer by Feature Decorationcited by 115
10 more
  1. 2016Combining Markov Random Fields and Convolutional Neural Networks for Image Synthesiscited by 102
  2. 2017StyleBank: An Explicit Representation for Neural Image Style Transfercited by 89
  3. 2020Arbitrary Video Style Transfer via Multi-Channel Correlationcited by 84
  4. 2017Diversified Texture Synthesis with Feed-forward Networkscited by 59
  5. 2019A Closed-form Solution to Universal Style Transfercited by 42
  6. 2019Diversified Arbitrary Style Transfer via Deep Feature Perturbationcited by 42
  7. 2021Domain-Aware Universal Style Transfercited by 40
  8. 2020Manifold Alignment for Semantically Aligned Style Transfercited by 34
  9. 2021PhotoWCT2: Compact Autoencoder for Photorealistic Style Transfer Resulting from Blockwise Training and Skip Connections of High-Frequency Residualscited by 30
  10. 2018Self-Contained Stylization via Steganography for Reverse and Serial Style Transfercited by 4

Built on by

Cited by year

Citing papers per year: 2022: 1; 2023: 1; 2024: 1.

3 papers.