Domain Generalization via Model-Agnostic Learning of Semantic Features

By Qi-Yan Dou, Daniel Coelho de Castro, Konstantinos Kamnitsas, Ben Glocker. 2019

812 citations recorded by the source corpus.

Generalization capability to unseen domains is crucial for machine learning models when deploying to real-world conditions. We investigate the challenging problem of domain generalization, i.e., training a model on multi-domain source data such that it can directly generalize to target domains with unknown statistics. We adopt a model-agnostic learning paradigm with gradient-based meta-train and meta-test procedures to expose the optimization to domain shift. Further, we introduce two complementary losses which explicitly regularize the semantic structure of the feature space. Globally, we align a derived soft confusion matrix to preserve general knowledge of inter-class relationships. Locally, we promote domain-independent class-specific cohesion and separation of sample features with a metric-learning component. The effectiveness of our method is demonstrated with new state-of-the-art results on two common object recognition benchmarks. Our method also shows consistent improvement on a medical image segmentation task.

Subfields: Artificial Intelligence, Computer Vision and Pattern Recognition, Radiology, Nuclear Medicine and Imaging

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

In Radiology, Nuclear Medicine and Imaging

Citations in the corpus

On Generalizing Beyond Domains in Cross-Domain Continual Learning (2022, background): “To name a few, the Contrastive Adaptation Network (CAN) for unsupervised domain adaptation [17], the Covariance Metric Networks (CovaMNet) for few-shot learning [22], the Model-Agnostic learning of Semantic Features (MASF) for standard domain generalization [6] and”
SWAD: Domain Generalization by Seeking Flat Minima (2021, neutral): “MASF [14] 82.7 - - - - - DMG [33] 83.4 - - - 43.6 - MetaReg [15] 83.6 - - - 43.6 - ER [12] 85.3 - - - - - pAdaIN [47] 85.4 - - - - - EISNet [48] 85.8 - - - - - DSON [30] 86.”
In Search of Lost Domain Generalization (2020, positive): “Dou et al. [2019] use a similar MAML strategy, together with two regularizers that encourage features from different domains to respect inter-class relationships, and be compactly clustered by class labels.”
WILDS: A Benchmark of in-the-Wild Distribution Shifts (2020, neutral): “Numerous methods have been proposed for domain generalization, including domain-invariant learning (Ganin et al., 2016; Sun and Saenko, 2016), invariant risk minimization (IRM) (Arjovsky et al., 2019), and meta-learning-based methods (Li et al., 2017; Dou et al., 2019).”
Unsupervised Bidirectional Cross-Modality Adaptation via Deeply Synergistic Image and Feature Alignment for Medical Image Segmentation (2020, neutral): “Plenty of adaptive approaches have been proposed with different strategies [11]–[13], [27]–[31].”
Adaptive Risk Minimization: A Meta-Learning Approach for Tackling Group Shift (2020, methodology): “Our method also resembles prior meta-learning methods for domain generalization [37, 14], which studies shift at the level of test domains.”
Domain Generalization via Gradient Surgery (2021, methodology): “Instead of domain-invariant feature representations, other methods propose extracting domain-agnostic parameters to address generalization through max-margin linear models [11], low-rank parametrized CNNs [15] and meta-learning [16, 5].”
Domain Generalization with MixStyle (2021, background): “Domain generalization (DG) aims to address such a problem (Zhou et al., 2021; Blanchard et al., 2011; Muandet et al., 2013; Li et al., 2018a; Zhou et al., 2020b; Balaji et al., 2018; Dou et al., 2019; Carlucci et al., 2019).”
Learning to Optimize Domain Specific Normalization for Domain Generalization (2019, methodology): “The proposed algorithm is compared with several existing methods, which include JiGen [4], D-SAM [6], Epi-FCR [14], MetaReg [2], MASF [7], and MMLD [18].”
Learning to Diversify for Single Domain Generalization (2021, methodology): “The first category,including DSN [4], Fusion [30], MetaReg [2], Epi- FCR [25], MASF [10] and DMG [6], requires domain identifications in the training phase.”
Orthogonal Projection Loss (2021, neutral): “In particular, [9] explores inter-class separation and intra-class clustering through contrastive and triplet loss functions.”
A Fourier-based Framework for Domain Generalization (2021, methodology): “Compared with the SOTA, our FACT clearly beats the methods based on adversarial data augmentation or meta learning, including the latest MASF [4], DDAIG [50] and L2A-OT [51], yet FACT enjoys an efficient training process without any additional adversarial or episodic training steps.”
Cross-Domain First Person Audio-Visual Action Recognition through Relative Norm Alignment (2021, methodology): “Previous approaches in DG are mostly designed for image data [6, 56, 30, 17, 31, 5] and are divided in feature-based and data-based methods.”
Reducing Domain Gap via Style-Agnostic Networks (2019, methodology): “Our framework seamlessly extends to single-source DG where only a single training domain is provided, because it does not require domain labels nor multiple source domains (which are necessary in the majority of DG methods [30,9,31,8]).”
RobustNet: Improving Domain Generalization in Urban-Scene Segmentation via Instance Selective Whitening (2021, background): “Various approaches such as meta-learning [27, 2, 28, 33], adversarial training [29, 32, 46], autoencoder [15, 29], metric learning [10, 39], data augmentation [64, 16, 67] have been proposed to learn domain-agnostic feature representations.”
Towards Stable Test-Time Adaptation in Dynamic Wild World (2023, methodology): “…shifts has been studied in a number of works at training time , including domain generalization (Shankar et al., 2018; Li et al., 2018a; Dou et al., 2019), increasing the training dataset size (Orhan, 2019), various data augmentation techniques (Lim et al., 2019; Hendrycks et al.,…”
Unleashing the Power of Contrastive Self-Supervised Visual Models via Contrast-Regularized Fine-Tuning (2021, background): “[12] Qi Dou, Daniel C Castro, Konstantinos Kamnitsas, and Ben Glocker.”