A Fourier-based Framework for Domain Generalization

By Qinwei Xu, Ruipeng Zhang, Ya Zhang, Yanfeng Wang, Qi Tian. Computer Vision and Pattern Recognition, 2021

641 citations recorded by the source corpus.

Modern deep neural networks suffer from performance degradation when evaluated on testing data under different distributions from training data. Domain generalization aims at tackling this problem by learning transferable knowledge from multiple source domains in order to generalize to unseen target domains. This paper introduces a novel Fourier-based perspective for domain generalization. The main assumption is that the Fourier phase information contains high-level semantics and is not easily affected by domain shifts. To force the model to capture phase information, we develop a novel Fourier-based data augmentation strategy called amplitude mix which linearly interpolates between the amplitude spectrums of two images. A dual-formed consistency loss called co-teacher regularization is further introduced between the predictions induced from original and augmented images. Extensive experiments on three benchmarks have demonstrated that the proposed method is able to achieve state-of-the-arts performance for domain generalization.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Mechanical Engineering

View Paper: https://www.semanticscholar.org/paper/5b27e6d2fa2705730b2af8ad7652b2bfaa0b80a0

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Artificial Intelligence

1. Tao Qin
2. Peng Cui
3. Renzhe Xu
4. Jindong Wang
5. Yidong Ouyang
6. Yue He
7. Jiashuo Liu
8. Chang Liu

In Computer Vision and Pattern Recognition

1. Peng Cui
2. Tao Qin
3. Renzhe Xu
4. Jindong Wang
5. Yidong Ouyang
6. Yue He
7. Jiashuo Liu
8. Chang Liu

Citations in the corpus

• Towards Out-Of-Distribution Generalization: A Survey (2021, background): “In addition to domain adversarial learning, numerous studies [248, 146, 169, 168, 167, 269] propose the alignment of features to cultivate domain invariant representations.”
• Generalizing to Unseen Domains: A Survey on Domain Generalization (2021, methodology): “Recently, there are several methods using Mixup for DG, by either performing Mixup in the original space [47, 48, 54] to generate new samples; or in the feature space [49, 148, 55] which does not explicitly generate raw training samples.”