Domain Generalization with MixStyle

By Kaiyang Zhou, Yongxin Yang, Yu Qiao, Tao A. Xiang. 2021

1,040 citations recorded by the source corpus.

Though convolutional neural networks (CNNs) have demonstrated remarkable ability in learning discriminative features, they often generalize poorly to unseen domains. Domain generalization aims to address this problem by learning from a set of source domains a model that is generalizable to any unseen domain. In this paper, a novel approach is proposed based on probabilistically mixing instance-level feature statistics of training samples across source domains. Our method, termed MixStyle, is motivated by the observation that visual domain is closely related to image style (e.g., photo vs.~sketch images). Such style information is captured by the bottom layers of a CNN where our proposed style-mixing takes place. Mixing styles of training instances results in novel domains being synthesized implicitly, which increase the domain diversity of the source domains, and hence the generalizability of the trained model. MixStyle fits into mini-batch training perfectly and is extremely easy to implement. The effectiveness of MixStyle is demonstrated on a wide range of tasks including category classification, instance retrieval and reinforcement learning.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition

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

In Computer Vision and Pattern Recognition

Citations in the corpus

On Generalizing Beyond Domains in Cross-Domain Continual Learning (2022, methodology): “[40] propose a smooth style transfer applied to the feature statistics.”
SWAD: Domain Generalization by Seeking Flat Minima (2021, methodology): “For Mixstyle [17], we insert Mixstyle block with domain label after the 1st, 2nd, and 3rd residual blocks with α = 0 .”
Generalizing to Unseen Domains: A Survey on Domain Generalization (2021, methodology): “[54] to generate new samples; or in the feature space [49], [55], [148] which does not explicitly generate raw training samples.”
Domain Generalization: A Survey (2021, methodology): “cation [23], [35], but also in other domains such as speech recognition [40], natural language processing [34], medical imaging [41], [42], and reinforcement learning [23].”
How Neural Networks Extrapolate: From Feedforward to Graph Neural Networks (2020, background): “Invariant models aim to learn features that respect specific invariances across multiple training distributions (Arjovsky et al., 2019; Rojas-Carulla et al., 2018; Zhou et al., 2021).”
GenerSpeech: Towards Style Transfer for Generalizable Out-Of-Domain Text-to-Speech Synthesis (2022, methodology): “Leveraging recent progress on domain generalization (Zhou et al., 2020), in this work, we design the Mix-Style Layer Normalization for regularizing TTS model training by perturbing the style information in training samples:”