Domain Generalization via Frequency-domain-based Feature Disentanglement and Interaction

Wang, Jingye; Du, Ruoyi; Chang, Dongliang; Liang, Kongming; Ma, Zhanyu

doi:10.1145/3503161.3548267

Cited by 19 publications

(3 citation statements)

References 31 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In [48], the amplitude swap (AS) and amplitude mix (AM) strategies are introduced for data augmentation, the former is exactly the same as the spectral transfer [50] and the latter is to mix amplitude of source and target domain. Similarly, [47] implements data augmentation by applying multiplicative and additive Gaussian noises to both of the amplitude and phase of the source domain. Most frequencybased works have focused on input-level data augmentation.…”

Section: Related Workmentioning

confidence: 99%

See 1 more Smart Citation

Decompose, Adjust, Compose: Effective Normalization by Playing with Frequency for Domain Generalization

Lee¹,

Bae²,

Kim³

2023

Preprint

View full text Add to dashboard Cite

Domain generalization (DG) is a principal task to evaluate the robustness of computer vision models. Many previous studies have used normalization for DG. In normalization, statistics and normalized features are regarded as style and content, respectively. However, it has a content variation problem when removing style because the boundary between content and style is unclear. This study addresses this problem from the frequency domain perspective, where amplitude and phase are considered as style and content, respectively. First, we verify the quantitative phase variation of normalization through the mathematical derivation of the Fourier transform formula. Then, based on this, we propose a novel normalization method, P CN orm, which eliminates style only as the preserving content through spectral decomposition. Furthermore, we propose advanced P CN orm variants, CCN orm and SCN orm, which adjust the degrees of variations in content and style, respectively. Thus, they can learn domain-agnostic representations for DG. With the normalization methods, we propose ResNet-variant models, DAC-P and DAC-SC, which are robust to the domain gap. The proposed models outperform other recent DG methods. The DAC-SC achieves an average state-of-the-art performance of 65.

show abstract

Section: Related Workmentioning

confidence: 99%

“…Moreover, another method for style-based DG [6,45,47,48] is the frequency domain-based method. Input images are decomposed into amplitude and phase using the Fourier transform (FT) [4].…”

Section: Introductionmentioning

confidence: 99%

Decompose, Adjust, Compose: Effective Normalization by Playing with Frequency for Domain Generalization

Lee¹,

Bae²,

Kim³

2023

Preprint

View full text Add to dashboard Cite

show abstract

“…Frequency information has been widely used in convolutional neural networks for improving performance [7]- [19] pruning and network compression [4]- [6], [20]- [30], or increasing the detection accuracy [31]- [33]. For instance, Wang et al [18] represented object edges and smooth structures using high and low-frequency information, respectively. Mi et al [17] split channel recognition networks into frequency domains, while Rippel et al [34] proposed a fully spectral representation of network parameters.…”

Section: Related Workmentioning

confidence: 99%

Frequency Regularization: Reducing Information Redundancy in Convolutional Neural Networks

Zhao,

Dong,

Zhang

et al. 2023

IEEE Access

View full text Add to dashboard Cite

Convolutional neural networks have demonstrated impressive results in many computer vision tasks. However, the increasing size of these networks raises concerns about the information overload resulting from the large number of network parameters. In this paper, we propose Frequency Regularization to restrict the non-zero elements of the network parameters in the frequency domain. The proposed approach operates at the tensor level, and can be applied to almost all network architectures. Specifically, the tensors of parameters are maintained in the frequency domain, where high-frequency components can be eliminated by zigzag setting tensor elements to zero. Then, the inverse discrete cosine transform (IDCT) is used to reconstruct the spatial tensors for matrix operations during network training. Since high-frequency components of images are known to be less critical, a large proportion of these parameters can be set to zero when networks are trained with the proposed frequency regularization. Comprehensive evaluations on various state-of-the-art network architectures, including LeNet, Alexnet, VGG, Resnet, ViT, UNet, GAN, and VAE, demonstrate the effectiveness of the proposed frequency regularization. For a very small accuracy decrease (less than 2%), a LeNet5 with 0.4M parameters can be represented by only 776 float16 numbers (over 1100× reduction), and a UNet with 34M parameters can be represented by only 759 float16 numbers (over 80000× reduction). In particular, the original size of the UNet model is reduced from 366 Mb to 4.5 Kb.

show abstract

Style Augmentation and Domain-Aware Parametric Contrastive Learning for Domain Generalization

Zhang

et al. 2023

Lecture Notes in Computer Science

View full text Add to dashboard Cite

Domain Generalization via Frequency-domain-based Feature Disentanglement and Interaction

Cited by 19 publications

References 31 publications

Decompose, Adjust, Compose: Effective Normalization by Playing with Frequency for Domain Generalization

Decompose, Adjust, Compose: Effective Normalization by Playing with Frequency for Domain Generalization

Frequency Regularization: Reducing Information Redundancy in Convolutional Neural Networks

Style Augmentation and Domain-Aware Parametric Contrastive Learning for Domain Generalization

Contact Info

Product

Resources

About