Unseen Target Domain Research Articles

A unified Personalized Federated Learning framework ensuring Domain Generalization

Personalized Federated Learning (pFL) allows for the development of customized models for personalized information from multiple distributed domains. In real-world scenarios, some testing data may originate from new target domains (unseen domains) outside of the federated network, resulting in another learning task called Federated Domain Generalization (FedDG). In this paper, we aim to tackle the new problem, named Personalized Federated Domain Generalization (pFedDG), which not only protects the personalization but also obtains a general model for unseen target domains. We observe that pFL and FedDG objectives can conflict, posing challenges in addressing both objectives simultaneously. To sufficiently moderate the conflict, we develop a unified framework, named Personalized Federated Decoupled Representation (pFedDR), which decouples the two objectives using two separate loss functions simultaneously and uses an integrated predictor to serve both two learning tasks. Specifically, the framework decouples domain-sensitive layers linked to different representations and design an entropy increase loss to encourage the separation of two representations to achieve the pFedDG. Extensive experiments show that our pFedDR method achieves state-of-the-art performance for both tasks while incurring almost no increase in communication cost. Code is available at https://github.com/CSU-YL/pFedDR.

Learning a generalizable re-identification model from unlabelled data with domain-agnostic expert

In response to real-world scenarios, the domain generalization (DG) problem has spurred considerable research in person re-identification (ReID). This challenge arises when the target domain, which is significantly different from the source domains, remains unknown. However, the performance of current DG ReID relies heavily on labor-intensive source domain annotations. Considering the potential of unlabeled data, we investigate unsupervised domain generalization (UDG) in ReID. Our goal is to create a model that can generalize from unlabeled source domains to semantically retrieve images in an unseen target domain. To address this, we propose a new approach that trains a domain-agnostic expert (DaE) for unsupervised domain-generalizable person ReID. This involves independently training multiple experts to account for label space inconsistencies between source domains. At the same time, the DaE captures domain-generalizable information for testing. Our experiments demonstrate the effectiveness of this method for learning generalizable features under the UDG setting. The results demonstrate the superiority of our method over state-of-the-art techniques. We will make our code and models available for public use.

Adaptive wavelet-VNet for single-sample test time adaptation in medical image segmentation.

In medical image segmentation, a domain gap often exists between training and testing datasets due to different scanners or imaging protocols, which leads to performance degradation in deep learning-based segmentation models. Given the high cost of manual labeling and the need for privacy protection, it is often challenging to annotate the testing (target) domain data for model fine-tuning or to collect data from different domains to train domain generalization models. Therefore, using only unlabeled target domain data for test-time adaptation (TTA) presents a more practical but challenging solution. To improve the segmentation accuracy of deep learning-based models on unseen datasets, and especially to enhance the efficiency and stability of TTA for individual samples from heterogeneous domains. In this study, we proposed to dynamically adapt a wavelet-VNet (WaVNet) to unseen target domains with a hybrid objective function, based on each unlabeled test sample during the test time. We embedded multiscale wavelet coefficients into a V-Net encoder and adaptively adjusted the spatial and spectral features according to the input, and the model parameters were optimized by three loss functions. We integrated a shape-aware loss to focus on the foreground segmentations, a Refine loss to correct the incomplete and noisy segmentations caused by domain shifts, and an entropy loss to promote the global consistency of the segmentations. We evaluated the proposed method on multidomain liver and prostate segmentation datasets to assess its advantages over other TTA methods. For the source domain model training of the liver dataset, we used 15 3D MR image samples for training and 5 for validation. Correspondingly, for the prostate dataset, we used 22 3D MR image samples for training and 7 for validation. In the target domain, we used a single 3D MR image sample for adaptation and testing. The total number of testing samples is 60 in the liver dataset (for 3 different domains) and 116 in the prostate dataset (for 6 different domains). The proposed method showed the highest segmentation accuracy among all methods, achieving a mean(±SD) Dice coefficient (DSC) of 78.10±5.23% and a mean 95th Hausdorff distance (HD95) of 15.52±5.84mm on the liver dataset; and a mean DSC of 80.02±3.89% and a mean HD95 of 9.18±3.47mm on the prostate dataset. The DSC is 11.67% (in absolute terms) and 15.27% higher than that of the baseline (no adaptation) method, for the liver and the prostate datasets, respectively. The proposed adaptive WaVNet enhanced the image segmentation accuracy from unseen domains during the test time via unsupervised learning and multi-objective optimization. It can benefit clinical applications where data are scarce or with changing data distributions, including online adaptive radiotherapy. The code will be released at: https://github.com/sanny1226/WaVNet.

Domain-augmented meta ensemble learning for mechanical fault diagnosis from heterogeneous source domains to unseen target domains

Existing domain generalization (DG) fault diagnosis methods primarily use adversarial training to reduce shifts between source domains and learn domain-invariant features. However, such features are difficult to learn when dealing with substantial shifts between source domains. In addition, focusing only on domain-invariant and ignoring domain-specific information is not conducive to improving model generalization. Furthermore, these methods are typically developed under an assumption that source domains share the same label space. In situations where the source domains exhibit heterogeneity with inconsistent labels, as explored in this paper, aligning the source domains becomes more difficult due to severe class mismatches. To address the above challenges, this paper proposes a method named domain-augmented meta ensemble learning. Specifically, Dirichlet CutMix is developed to compensate for missing classes in a source domain by utilizing knowledge from other source domains. Moreover, a training strategy summarized as “learn to generalize to unseen domains through collaborative ensemble” is designed to balance the learning of domain-specific features with the ability to generalize across domains. The proposed method is applied to the bearing and gearbox fault diagnosis, and experimental results demonstrate the excellent generalization of the method from heterogeneous source domains to unseen target domains.

A domain feature decoupling network for rotating machinery fault diagnosis under unseen operating conditions

Operating conditions reflect the mission evolution of rotating machinery in specific application scenarios. The monitoring data under different operating conditions exhibit disparate statistic distributions in terms of amplitude and frequency properties. In the modern operation and maintenance, domain adaptation approaches, which align the feature distributions of source and target domains, are commonly employed to address cross-domain diagnosis challenges. However, the availability of the target domain is serendipitous rather than guaranteed. To meet the engineering requirements, a domain feature decoupling network (DFDN) is proposed to achieve the fault diagnosis of rotating machinery under unseen operating conditions. The core concept of DFDN is to decouple the extracted features into domain-related features and fault-related features. These features are then adaptively integrated into domain-invariant features using a significance fusion method, thereby enabling the feature learning and knowledge transfer from seen source domains to unseen target domains. Furthermore, a self-learning joint optimization strategy is developed for DFDN to facilitate the efficient acquisition of generalized diagnosis knowledge. Two bearing datasets from different test rigs are considered to assess the domain generalization performance of DFDN. The experimental results demonstrate the superiority and potential of this method in coping with zero-shot fault diagnosis under unseen operating conditions.

Mask-Shift-Inference: A novel paradigm for domain generalization

Domain Generalization (DG) focuses on the Out-Of-Distribution (OOD) generalization, which is able to learn a robust model that generalizes the knowledge acquired from the source domain to the unseen target domain. However, due to the existence of the domain shift, domain-invariant representation learning is challenging. Guided by fine-grained knowledge, we propose a novel paradigm Mask-Shift-Inference (MSI) for DG based on the architecture of Convolutional Neural Networks (CNN). Different from relying on a series of constraints and assumptions for model optimization, this paradigm novelly shifts the focus to feature channels in the latent space for domain-invariant representation learning. We put forward a two-branch working mode of a main module and multiple domain-specific sub-modules. The latter can only achieve good prediction performance in its own specific domain but poor predictions in other source domains, which provides the main module with the fine-grained knowledge guidance and contributes to the improvement of the cognitive ability of MSI. Firstly, during the forward propagation of the main module, the proposed MSI accurately discards unstable channels based on spurious classifications varying across domains, which have domain-specific prediction limitations and are not conducive to generalization. In this process, a progressive scheme is adopted to adaptively increase the masking ratio according to the training progress to further reduce the risk of overfitting. Subsequently, our paradigm enters the compatible shifting stage before the formal prediction. Based on maximizing semantic retention, we implement the domain style matching and shifting through the simple transformation through Fourier transform, which can explicitly and safely shift the target domain back to the source domain whose style is closest to it, requiring no additional model updates and reducing the domain gap. Eventually, the paradigm MSI enters the formal inference stage. The updated target domain is predicted in the main module trained in the previous stage with the benefit of familiar knowledge from the nearest source domain masking scheme. Our paradigm is logically progressive, which can intuitively exclude the confounding influence of domain-specific spurious information along with mitigating domain shifts and implicitly perform semantically invariant representation learning, achieving robust OOD generalization. Extensive experimental results on PACS, VLCS, Office-Home and DomainNet datasets verify the superiority and effectiveness of the proposed method.

Domain generalization for semantic segmentation: a survey

Deep neural networks (DNNs) have proven explicit contributions in making autonomous driving cars and related tasks such as semantic segmentation, motion tracking, object detection, sensor fusion, and planning. However, in challenging situations, DNNs are not generalizable because of the inherent domain shift due to the nature of training under the i.i.d. assumption. The goal of semantic segmentation is to preserve information from a given image into multiple meaningful categories for visual understanding. Particularly for semantic segmentation, pixel-wise annotation is extremely costly and not always feasible. Domain generalization for semantic segmentation aims to learn pixel-level semantic labels from multiple source domains and generalize to predict pixel-level semantic labels on multiple unseen target domains. In this survey, for the first time, we present a comprehensive review of DG for semantic segmentation. we present a comprehensive summary of recent works related to domain generalization in semantic segmentation, which establishes the importance of generalizing to new environments of segmentation models. Although domain adaptation has gained more attention in segmentation tasks than domain generalization, it is still worth unveiling new trends that are adopted from domain generalization methods in semantic segmentation. We cover most of the recent and dominant DG methods in the context of semantic segmentation and also provide some other related applications. We conclude this survey by highlighting the future directions in this area.

Implicit Sharpness-Aware Minimization for Domain Generalization

Domain generalization (DG) aims to learn knowledge from multiple related domains to achieve a robust generalization performance in unseen target domains, which is an effective approach to mitigate domain shift in remote sensing image classification. Although the sharpness-aware minimization (SAM) method enhances DG capability and improves remote sensing image classification performance by promoting the convergence of the loss minimum to a flatter loss surface, the perturbation loss (maximum loss within the neighborhood of a local minimum) of SAM fails to accurately measure the true sharpness of the loss landscape. Furthermore, its variants often overlook gradient conflicts, thereby limiting further improvement in DG performance. In this paper, we introduce implicit sharpness-aware minimization (ISAM), a novel method that addresses the deficiencies of SAM and mitigates gradient conflicts. Specifically, we demonstrate that the discrepancy in training loss during gradient ascent or descent serves as an equivalent measure of the dominant eigenvalue of the Hessian matrix. This discrepancy provides a reliable measure for sharpness. ISAM effectively reduces sharpness and mitigates potential conflicts between gradients by implicitly minimizing the discrepancy between training losses while ensuring a sufficiently low minimum through minimizing perturbation loss. Extensive experiments and analyses demonstrate that ISAM significantly enhances the model’s generalization ability on remote sensing and DG datasets, outperforming existing state-of-the-art methods.

Source domain prior-assisted segment anything model for single domain generalization in medical image segmentation

Deep learning-based medical image segmentation models often suffer from performance degradation across domains due to domain discrepancies arising from data collected by various healthcare centers. Recent advancements, particularly the Segment Anything Model (SAM), have shown promising generalization abilities in this field, inspiring the development of several SAM-based approaches to address domain discrepancy issue. Nevertheless, these methods seldom exploit the full potential of the rich source domain knowledge to improve segmentation accuracy in unseen target domains. In this paper, we propose a source domain prior-assisted module for generalizable medical image segmentation based on SAM. Specifically, we store diverse features of the source domain data in a memory bank. When applying the model to the target domain, the features of the target domain first match with the features in the memory bank to obtain invaluable prior information. This strategy enables the model to utilize the prior information from the source domain to adapt to the target domain. We validate the proposed method on two widely used medical image segmentation tasks across multiple domains and achieve state-of-the-art performance.

Domain generalized person reidentification based on skewness regularity of higher-order statistics

The goal of domain-generalized person reidentification (DG-ReID) is to train a model in the source domain and apply it directly to unknown target domains for specific pedestrian retrieval. Existing methods rely primarily on low-order statistics (such as the mean, standard deviation, or variance), thereby ensuring the stability of the source domain data distribution for model training. However, such methods underperform when the data follow a non-Gaussian distribution, thereby reducing the generalization ability of the model on unseen target domains. To address this issue, this study proposes an instance normalization-based skewness regularity (INSR) framework that uses high-order statistics (skewness and high-order moments) to measure the skewness and regularity of the data distribution. Such measures allow further learning of the morphological features (skewness degree, trait of data near the mean, etc.) of complex data distributions for DG-ReID. Specifically, the proposed framework first extracts the skewness and third-order moments from the source domains, which provide more features (high-order moments, variance, etc.) to characterize the data distribution. Subsequently, a batch normalization-like operation was implemented to project the data into a new feature space with zero mean and unit variance, enhancing model adaption and accuracy. Extensive experiments were conducted on small-scale (VIPeR, PRID, GRID, and i-LIDS) and large-scale (Market-1501, DukeMTMC-reID, CUHK03, MSMT17) public datasets using two different protocols, demonstrating that the proposed INSR framework significantly outperforms other state-of-the-art counterparts for DG-ReID.

Low-resource entity resolution with domain generalization and active learning

Entity Resolution (ER), a fundamental task in data cleaning and integration, is critical in various fields such as healthcare, e-commerce, and social networks. Traditional ER methods are constrained by the need for substantial labeled samples and the challenge of generalization to unseen domains. To address the low-resource challenge in ER, a novel two-phase framework is proposed. Initially, we introduce the Domain Generalization Entity Resolution (DGER) framework, combining domain adversarial learning and simulated target learning to improve the generalization performance on unseen target domains. Subsequently, to further adapt to the target dataset, we present a novel active learning approach called Domain-Aware Uncertainty Active Learning (DUAL), for fine-tuning the DGER model with minimal annotation cost. DUAL manually annotates target domain samples that are highly uncertain and exhibit high divergence from the source, while assigning pseudo-labels to high-confidence samples. Experimental results on multiple real-world datasets demonstrate that our framework outperforms traditional ER methods in generalizing to unseen domains. Specifically, our DGER method outperforms the best-performing ER baseline in each task, achieving an average F1 score improvement of 9.02% across eight different test tasks. Moreover, within a limited annotation budget during the active learning phase, our DUAL fine-tuning strategy for the ER model outperforms uncertainty-based active learning techniques.

A unified deep semantic expansion framework for domain-generalized person re-identification

Supervised Person Re-identification (Person ReID) methods have achieved excellent performance when training and testing within one camera network. However, they usually suffer from considerable performance degradation when applied to different camera systems. In recent years, many Domain Adaptation Person ReID methods have been proposed, achieving impressive performance without requiring labeled data from the target domain. However, these approaches still need the unlabeled data of the target domain during the training process, making them impractical in many real-world scenarios. Our work focuses on the more practical Domain Generalized Person Re-identification (DG-ReID) problem. Given one or more source domains, it aims to learn a generalized model that can be applied to unseen target domains. One promising research direction in DG-ReID is the use of implicit deep semantic feature expansion, and our previous method, Domain Embedding Expansion (DEX), is one such example that achieves powerful results in DG-ReID. However, in this work we show that DEX and other similar implicit deep semantic feature expansion methods, due to limitations in their proposed loss function, fail to reach their full potential on large evaluation benchmarks as they have a tendency to saturate too early. Leveraging on this analysis, we propose Unified Deep Semantic Expansion, our novel framework that unifies implicit and explicit semantic feature expansion techniques in a single framework to mitigate this early over-fitting and achieve a new state-of-the-art (SOTA) in all DG-ReID benchmarks. Further, we apply our method on more general image retrieval tasks, also surpassing the current SOTA in all of these benchmarks by wide margins.

Learning Robust Shape Regularization for Generalizable Medical Image Segmentation.

Generalizable medical image segmentation enables models to generalize to unseen target domains under domain shift issues. Recent progress demonstrates that the shape of the segmentation objective, with its high consistency and robustness across domains, can serve as a reliable regularization to aid the model for better cross-domain performance, where existing methods typically seek a shared framework to render segmentation maps and shape prior concurrently. However, due to the inherent texture and style preference of modern deep neural networks, the edge or silhouette of the extracted shape will inevitably be undermined by those domain-specific texture and style interferences of medical images under domain shifts. To address this limitation, we devise a novel framework with a separation between the shape regularization and the segmentation map. Specifically, we first customize a novel whitening transform-based probabilistic shape regularization extractor namely WT-PSE to suppress undesirable domain-specific texture and style interferences, leading to more robust and high-quality shape representations. Second, we deliver a Wasserstein distance-guided knowledge distillation scheme to help the WT-PSE to achieve more flexible shape extraction during the inference phase. Finally, by incorporating domain knowledge of medical images, we propose a novel instance-domain whitening transform method to facilitate a more stable training process with improved performance. Experiments demonstrate the performance of our proposed method on both multi-domain and single-domain generalization.

Domain Generalization via Ensemble Stacking for Face Presentation Attack Detection

AbstractFace presentation attack detection (PAD) plays a pivotal role in securing face recognition systems against spoofing attacks. Although great progress has been made in designing face PAD methods, developing a model that can generalize well to unseen test domains remains a significant challenge. Moreover, due to the different types of spoofing attacks, creating a dataset with a sufficient number of samples for training deep neural networks is a laborious task. This work proposes a comprehensive solution that combines synthetic data generation and deep ensemble learning to enhance the generalization capabilities of face PAD. Specifically, synthetic data is generated by blending a static image with spatiotemporal-encoded images using alpha composition and video distillation. In this way, we simulate motion blur with varying alpha values, thereby generating diverse subsets of synthetic data that contribute to a more enriched training set. Furthermore, multiple base models are trained on each subset of synthetic data using stacked ensemble learning. This allows the models to learn complementary features and representations from different synthetic subsets. The meta-features generated by the base models are used as input for a new model called the meta-model. The latter combines the predictions from the base models, leveraging their complementary information to better handle unseen target domains and enhance overall performance. Experimental results from seven datasets—WMCA, CASIA-SURF, OULU-NPU, CASIA-MFSD, Replay-Attack, MSU-MFSD, and SiW-Mv2—highlight the potential to enhance presentation attack detection by using large-scale synthetic data and a stacking-based ensemble approach.

Restoration towards decomposition: A simple approach for domain generalization

Domain generalization (DG) aims to train a model capable of generalizing to unseen target domains by utilizing data from multiple disjoint source domains. Previous domain generalization methods primarily utilize data augmentation and domain-invariant feature learning. However, since the target domain is unknown, data augmentation methods struggle to generate images or features that closely resemble the target domain. In addition, existing domain-invariant feature learning methods are incapable of achieving complete decoupling, resulting in misalignment of feature distribution between the source and target domains. Hence, in this work, we propose a new perspective to address domain generalization by focusing on learning the ability to transform the target domain visual representations into those of the source domain. If the style of the unknown target domain can be transformed into the style of the known source domain, the adverse effects of domain shift can be effectively mitigated. Following this line of thought, we have developed a meta-learning task named “feature restoration” aimed at training a model to acquire this capability. Specifically, feature restoration involves replacing the domain-specific components within the visual representations of the target domain with those of the source domain. Experimental results on five DG benchmarks reveal that our method can achieve state-of-the-art performance. Ablation studies and visualization results further demonstrate the rationality and effectiveness of our design.

FSN: Feature Shift Network for Load-Domain (LD) Domain Generalization

Conventional deep learning methods for fault detection often assume that the training and the testing sets share the same fault domain spaces. However, some fault patterns are rare, and many real-world faults have not appeared in the training set. As a result, it is hard for the trained model to achieve desirable performance on the testing set. In this paper, we introduce a novel domain generalization, Load-Domain (LD) domain generalization, which is based on the analysis of the Case Western Reserve University (CWRU) bearing dataset and takes advantage of the physical information of this dataset. For this scenario, we propose a feature shift model called Feature Shift Network (FSN). FSN is trained for feature shift on adjacent source domains and finally shifts target domain features into adjacent source domain feature space to achieve the purpose of domain generalization. Furthermore, through the hybrid classification method, the generalization performance of the model on unseen target domains is effectively improved. The results on the CWRU bearing dataset demonstrate that FSN is better than the existing models in the LD domain generalization. Furthermore, we have another test on the rotated MNIST, which also shows FSN can achieve the best performance.

Unbiased Semantic Representation Learning Based on Causal Disentanglement for Domain Generalization

Domain generalization primarily mitigates domain shift among multiple source domains, generalizing the trained model to an unseen target domain. However, the spurious correlation usually caused by context prior (e.g., background) makes it challenging to get rid of the domain shift. Therefore, it is critical to model the intrinsic causal mechanism. The existing domain generalization methods only attend to disentangle the semantic and context-related features by modeling the causation between input and labels, which totally ignores the unidentifiable but important confounders. In this article, a Causal Disentangled Intervention Model (CDIM) is proposed for the first time, to the best of our knowledge, to construct confounders via causal intervention. Specifically, a generative model is employed to disentangle the semantic and context-related features. The contextual information of each domain from generative model can be considered as a confounder layer, and the center of all context-related features is utilized for fine-grained hierarchical modeling of confounders. Then the semantic and confounding features from each layer are combined to train an unbiased classifier, which exhibits both transferability and robustness across an unknown distribution domain. CDIM is evaluated on three widely recognized benchmark datasets, namely, Digit-DG, PACS, and NICO, through extensive ablation studies. The experimental results clearly demonstrate that the proposed model achieves state-of-the-art performance.

Stereo sample generation‐based domain generalization network for stereo matching

AbstractRecently, deep learning‐based stereo matching has achieved great success. However, models trained on the source domain dataset encounter substantial performance degradation when directly tested on an unseen target domain dataset because of neglecting the generalization to out‐of‐distribution (OOD) stereo samples. This paper proposes a stereo sample generation‐based domain generalization network (SGDG‐Net) for stereo matching. Specifically, to expand the distribution span of training samples, OOD stereo samples are generated to assist training. To effectively generate OOD left samples, a style transfer‐based generation mechanism is proposed to transmit perturbations to the source left samples. In addition, to generate the OOD right samples, a disparity‐assisted generation strategy is proposed by using disparity map labels as auxiliary information. Experimental results demonstrate that the proposed SGDG‐Net produces remarkable results on four benchmark datasets.

Disease-driven domain generalization for neuroimaging-based assessment of Alzheimer's disease.

Development of deep learning models to evaluate structural brain changes caused by cognitive impairment in MRI scans holds significant translational value. The efficacy of these models often encounters challenges due to variabilities arising from different data generation protocols, imaging equipment, radiological artifacts, and shifts in demographic distributions. Domain generalization (DG) techniques show promise in addressing these challenges by enabling the model to learn from one or more source domains and apply this knowledge to new, unseen target domains. Here we present a framework that utilizes model interpretability to enhance the generalizability of classification models across various cohorts. We used MRI scans and clinical diagnoses from four independent cohorts: Alzheimer's Disease Neuroimaging Initiative (ADNI, n = 1821), the Framingham Heart Study (FHS, n = 304), the Australian Imaging Biomarkers & Lifestyle Study of Ageing (AIBL, n = 661), and the National Alzheimer's Coordinating Center (NACC, n = 4647). With this data, we trained a deep neural network to focus on areas of the brain identified as relevant to the disease for model training. Our approach involved training a classifier to differentiate between structural neurodegeneration in individuals with normal cognition (NC), mild cognitive impairment (MCI), and dementia due to Alzheimer's disease (AD). This was achieved by aligning class-wise attention with a unified visual saliency prior, which was computed offline for each class using all the training data. Our method not only competes with state-of-the-art approaches but also shows improved correlation with postmortem histology. This alignment with the gold standard evidence is a significant step towards validating the effectiveness of DG frameworks, paving the way for their broader application in the field.

Invariant Content Representation for Generalizable Medical Image Segmentation.

Domain generalization (DG) for medical image segmentation due to privacy preservation prefers learning from a single-source domain and expects good robustness on unseen target domains. To achieve this goal, previous methods mainly use data augmentation to expand the distribution of samples and learn invariant content from them. However, most of these methods commonly perform global augmentation, leading to limited augmented sample diversity. In addition, the style of the augmented image is more scattered than the source domain, which may cause the model to overfit the style of the source domain. To address the above issues, we propose an invariant content representation network (ICRN) to enhance the learning of invariant content and suppress the learning of variability styles. Specifically, we first design a gamma correction-based local style augmentation (LSA) to expand the distribution of samples by augmenting foreground and background styles, respectively. Then, based on the augmented samples, we introduce invariant content learning (ICL) to learn generalizable invariant content from both augmented and source-domain samples. Finally, we design domain-specific batch normalization (DSBN) based style adversarial learning (SAL) to suppress the learning of preferences for source-domain styles. Experimental results show that our proposed method improves by 8.74% and 11.33% in overall dice coefficient (Dice) and reduces 15.88 mm and 3.87 mm in overall average surface distance (ASD) on two publicly available cross-domain datasets, Fundus and Prostate, compared to the state-of-the-art DG methods.The code is available at https://github.com/ZMC-IIIM/ICRN-DG .