Domain-general Framework Research Articles

Unobtrusive High-Generalization Fall Detection: A Domain-Generalization Framework

Unobtrusive High-Generalization Fall Detection: A Domain-Generalization Framework

EEG-DG: A Multi-Source Domain Generalization Framework for Motor Imagery EEG Classification.

Motor imagery EEG classification plays a crucial role in non-invasive Brain-Computer Interface (BCI) research. However, the performance of classification is affected by the non-stationarity and individual variations of EEG signals. Simply pooling EEG data with different statistical distributions to train a classification model can severely degrade the generalization performance. To address this issue, the existing methods primarily focus on domain adaptation, which requires access to the test data during training. This is unrealistic and impractical in many EEG application scenarios. In this paper, we propose a novel multi-source domain generalization framework called EEG-DG, which leverages multiple source domains with different statistical distributions to build generalizable models on unseen target EEG data. We optimize both the marginal and conditional distributions to ensure the stability of the joint distribution across source domains and extend it to a multi-source domain generalization framework to achieve domain-invariant feature representation, thereby alleviating calibration efforts. Systematic experiments conducted on a simulative dataset, BCI competition IV 2a, 2b, and OpenBMI datasets, demonstrate the superiority and competitive performance of our proposed framework over other state-of-the-art methods. Specifically, EEG-DG achieves average classification accuracies of 81.79% and 87.12% on datasets IV-2a and IV-2b, respectively, and 78.37% and 76.94% for inter-session and inter-subject evaluations on dataset OpenBMI, which even outperforms some domain adaptation methods. Our code is available at https://github.com/zxchit2022/EEG-DG for evaluation.

A novel causal feature learning-based domain generalization framework for bearing fault diagnosis with a mixture of data from multiple working conditions and machines

Data distribution shift usually deteriorates the performance of traditional supervised learning-based fault diagnosis models. Such issue can be well addressed by domain generalization fault diagnosis methods even if the test data are unavailable in training stage. However, most of these methods are designed and validated only for a single device, and also require the domain labels being provided, which hampers their further application. Therefore, this paper focuses a more realistic scenario, where the source training set is a mixture of data from multiple operation conditions and devices with domain information unavailable. First, a causal analysis is made for the task of domain generalization fault diagnosis, and a structural causal model (SCM) that describes the data generation process is constructed based on the relationships among different variables. A new learning rule for domain generalization, named object-conditional domain invariant (Obj-DI) rule, is also elicited. Then, to achieve domain generalization fault diagnosis, a three-stage based method named LOODG is proposed. Potential domains are mined and samples are given domain labels in latent domain discovery stage. Samples belonging to same object but different domains are explored in object discovery stage. In Obj-DI based causal feature learning stage, generalizable causal features are learned to finally obtain a generalized model. Experiments are conducted on tasks designed from three datasets, and the results demonstrate the superiority of the proposed method.

Fault vibration model driven fault-aware domain generalization framework for bearing fault diagnosis

Deep learning methods can learn effective representations from the data, simplifying the fault diagnosis process and improving accuracy. However, the lack of data presents a significant challenge to effectively implementing fault diagnosis driven by deep learning. Simulation models can generate rich source domain data, which can be combined with domain adaptation methods to realize fault diagnosis tasks under sample scarcity. Nevertheless, there are some problems with existing simulation-driven fault diagnosis frameworks: (1) The choice of parameters for simulation models is usually empirical, which can lead to domain-specific components differences between the simulation data and the actual data. (2) It is difficult for the simulation models to generate the background noise of the real equipment, which will cause fault information to be covered, severely limiting the robustness of the fault diagnosis network. (3) Studies based on domain adaptation methods require a certain amount of target domain data for training, which is difficult to meet in real industrial production. To address these issues, this paper proposes a fault-aware domain generalization bearing fault diagnosis framework driven by fault vibration model, an analytical simulation model. Firstly, a spectral background estimation algorithm based on adaptive clutter separation (ACS) and maximum spectral envelope (MSE), termed ACS-MSE, is designed to estimate the potential resonance frequencies in the real system, guiding the parameters setting of the fault diagnosis vibration model and ensuring consistency between the simulated data and the actual data in the label space. Secondly, the normal actual signal is combined with the fault simulation signal to create a mixed signal that considers real background noise, and a fault-aware autoencoder is developed to extract the fault-related impact components covered by the background noise. Finally, a multi resonance frequencies domain generalization fault diagnosis network based on contrastive learning is designed to extract generalized fault features crossing different resonance frequencies and realize diagnosis tasks without real fault data. Experimental results on an open dataset and a custom dataset showed that the proposed method achieves higher accuracy compared to some advanced domain generalization and even domain adaptation methods. Ablation studies verify the importance and irreplaceability of the individual components of the proposed method.

CauDR: A causality-inspired domain generalization framework for fundus-based diabetic retinopathy grading

Diabetic retinopathy (DR) is the most common diabetic complication, which usually leads to retinal damage, vision loss, and even blindness. A computer-aided DR grading system has a significant impact on helping ophthalmologists with rapid screening and diagnosis. Recent advances in fundus photography have precipitated the development of novel retinal imaging cameras and their subsequent implementation in clinical practice. However, most deep learning-based algorithms for DR grading demonstrate limited generalization across domains. This inferior performance stems from variance in imaging protocols and devices inducing domain shifts. We posit that declining model performance between domains arises from learning spurious correlations in the data. Incorporating do-operations from causality analysis into model architectures may mitigate this issue and improve generalizability. Specifically, a novel universal structural causal model (SCM) was proposed to analyze spurious correlations in fundus imaging. Building on this, a causality-inspired diabetic retinopathy grading framework named CauDR was developed to eliminate spurious correlations and achieve more generalizable DR diagnostics. Furthermore, existing datasets were reorganized into 4DR benchmark for DG scenario. Results demonstrate the effectiveness and the state-of-the-art (SOTA) performance of CauDR. Diabetic retinopathy (DR) is the most common diabetic complication, which usually leads to retinal damage, vision loss, and even blindness. A computer-aided DR grading system has a significant impact on helping ophthalmologists with rapid screening and diagnosis. Recent advances in fundus photography have precipitated the development of novel retinal imaging cameras and their subsequent implementation in clinical practice. However, most deep learning-based algorithms for DR grading demonstrate limited generalization across domains. This inferior performance stems from variance in imaging protocols and devices inducing domain shifts. We posit that declining model performance between domains arises from learning spurious correlations in the data. Incorporating do-operations from causality analysis into model architectures may mitigate this issue and improve generalizability. Specifically, a novel universal structural causal model (SCM) was proposed to analyze spurious correlations in fundus imaging. Building on this, a causality-inspired diabetic retinopathy grading framework named CauDR was developed to eliminate spurious correlations and achieve more generalizable DR diagnostics. Furthermore, existing datasets were reorganized into 4DR benchmark for DG scenario. Results demonstrate the effectiveness and the state-of-the-art (SOTA) performance of CauDR.

Domain Generalizable Person Search Using Unreal Dataset

Collecting and labeling real datasets to train the person search networks not only requires a lot of time and effort, but also accompanies privacy issues. The weakly-supervised and unsupervised domain adaptation methods have been proposed to alleviate the labeling burden for target datasets, however, their generalization capability is limited. We introduce a novel person search method based on the domain generalization framework, that uses an automatically labeled unreal dataset only for training but is applicable to arbitrary unseen real datasets. To alleviate the domain gaps when transferring the knowledge from the unreal source dataset to the real target datasets, we estimate the fidelity of person instances which is then used to train the end-to-end network adaptively. Moreover, we devise a domain-invariant feature learning scheme to encourage the network to suppress the domain-related features. Experimental results demonstrate that the proposed method provides the competitive performance to existing person search methods even though it is applicable to arbitrary unseen datasets without any prior knowledge and re-training burdens.

Geometry-Guided Domain Generalization for Monocular 3D Object Detection

Monocular 3D object detection (M3OD) is important for autonomous driving. However, existing deep learning-based methods easily suffer from performance degradation in real-world scenarios due to the substantial domain gap between training and testing. M3OD's domain gaps are complex, including camera intrinsic parameters, extrinsic parameters, image appearance, etc. Existing works primarily focus on the domain gaps of camera intrinsic parameters, ignoring other key factors. Moreover, at the feature level, conventional domain invariant learning methods generally cause the negative transfer issue, due to the ignorance of dependency between geometry tasks and domains. To tackle these issues, in this paper, we propose MonoGDG, a geometry-guided domain generalization framework for M3OD, which effectively addresses the domain gap at both camera and feature levels. Specifically, MonoGDG consists of two major components. One is geometry-based image reprojection, which mitigates the impact of camera discrepancy by unifying intrinsic parameters, randomizing camera orientations, and unifying the field of view range. The other is geometry-dependent feature disentanglement, which overcomes the negative transfer problems by incorporating domain-shared and domain-specific features. Additionally, we leverage a depth-disentangled domain discriminator and a domain-aware geometry regression attention mechanism to account for the geometry-domain dependency. Extensive experiments on multiple autonomous driving benchmarks demonstrate that our method achieves state-of-the-art performance in domain generalization for M3OD.

Gaze from Origin: Learning for Generalized Gaze Estimation by Embedding the Gaze Frontalization Process

Gaze estimation aims to accurately estimate the direction or position at which a person is looking. With the development of deep learning techniques, a number of gaze estimation methods have been proposed and achieved state-of-the-art performance. However, these methods are limited to within-dataset settings, whose performance drops when tested on unseen datasets. We argue that this is caused by infinite and continuous gaze labels. To alleviate this problem, we propose using gaze frontalization as an auxiliary task to constrain gaze estimation. Based on this, we propose a novel gaze domain generalization framework named Gaze Frontalization-based Auxiliary Learning (GFAL) Framework which embeds the gaze frontalization process, i.e., guiding the feature so that the eyeball can rotate and look at the front (camera), without any target domain information during training. Experimental results show that our proposed framework is able to achieve state-of-the-art performance on gaze domain generalization task, which is competitive with or even superior to the SOTA gaze unsupervised domain adaptation methods.

Prompt-driven Latent Domain Generalization for Medical Image Classification.

Deep learning models for medical image analysis easily suffer from distribution shifts caused by dataset artifact bias, camera variations, differences in the imaging station, etc., leading to unreliable diagnoses in real-world clinical settings. Domain generalization (DG) methods, which aim to train models on multiple domains to perform well on unseen domains, offer a promising direction to solve the problem. However, existing DG methods assume domain labels of each image are available and accurate, which is typically feasible for only a limited number of medical datasets. To address these challenges, we propose a unified DG framework for medical image classification without relying on domain labels, called Prompt-driven Latent Domain Generalization (PLDG). PLDG consists of unsupervised domain discovery and prompt learning. This framework first discovers pseudo domain labels by clustering the bias-associated style features, then leverages collaborative domain prompts to guide a Vision Transformer to learn knowledge from discovered diverse domains. To facilitate cross-domain knowledge learning between different prompts, we introduce a domain prompt generator that enables knowledge sharing between domain prompts and a shared prompt. A domain mixup strategy is additionally employed for more flexible decision margins and mitigates the risk of incorrect domain assignments. Extensive experiments on three medical image classification tasks and one debiasing task demonstrate that our method can achieve comparable or even superior performance than conventional DG algorithms without relying on domain labels. Our code is publicly available at https://github.com/SiyuanYan1/PLDG/tree/main.

Meta-learning Based Domain Generalization Framework for Fault Diagnosis With Gradient Aligning and Semantic Matching

Intelligent fault diagnosis models have demonstrated a superior performance in industrial prognostics health management scenarios. However, these models may struggle to generalize in complicated industrial environments, when encountering new working conditions and handling low-resource and heterogeneous data. To cope with the aforementioned issues, we focus on constructing a universal training framework with domain generalization technique that will encourage fault diagnosis model to generalize well in unseen working conditions. Firstly, a model-agnostic meta-learning based training framework called Meta-GENE is proposed for homogeneous and heterogeneous domain generalization. Secondly, a gradient aligning algorithm is introduced in meta-learning framework to learn domain-invariant strategy for robust prediction in unseen working conditions. Thirdly, a semantic matching technique is proposed for utilizing heterogeneous data to alleviate low-resource problem. Our method has yielded excellent performance on the PHM09 fault diagnosis dataset and achieved superior results on a set of generalization tasks across various working conditions.

A federated distillation domain generalization framework for machinery fault diagnosis with data privacy

Federated learning is an emerging technology that enables multiple clients to cooperatively train an intelligent diagnostic model while preserving data privacy. However, federated diagnostic models still suffer from a performance drop when applied to entirely unseen clients outside the federation in practical deployments. To address this issue, a Federated Distillation Domain Generalization (FDDG) framework is proposed for machinery fault diagnosis. The core idea is to enable individual clients to access multi-client data distributions in a privacy-preserving manner and further explore domain invariance to enhance model generalization. A novel diagnostic knowledge-sharing mechanism is designed based on knowledge distillation, which equips multiple generators to augment fake data during the training of local models. Based on generated data and real data, a low-rank decomposition method is utilized to mine domain invariance, enhancing the model's ability to resist domain shift. Extensive experiments on two rotating machines demonstrate that the proposed FDDG achieves a 3% improvement in accuracy compared to state-of-the-art methods.

A Dual-Attention Deep Discriminative Domain Generalization Model for Hyperspectral Image Classification

Recently, hyperspectral image classification has made great progress with the development of convolutional neural networks. However, due to the challenges of distribution shifts and data redundancies, the classification accuracy is low. Some existing domain adaptation methods try to mitigate the distribution shifts by training source samples and some labeled target samples. However, in practice, labeled target domain samples are difficult or even impossible to obtain. To solve the above challenges, we propose a novel dual-attention deep discriminative domain generalization framework (DAD3GM) for cross-scene hyperspectral image classification without training the labeled target samples. In DAD3GM, we mainly design two blocks: dual-attention feature learning (DAFL) and deep discriminative feature learning (DDFL). DAFL is designed to extract spatial features by multi-scale self-attention and extract spectral features by multi-head external attention. DDFL is further designed to extract deep discriminative features by contrastive regularization and class discrimination regularization. The combination of DAFL and DDFL can effectively reduce the computational time and improve the generalization performance of DAD3GM. The proposed model achieves 84.25%, 83.53%, and 80.63% overall accuracy on the public Houston, Pavia, and GID benchmarks, respectively. Compared with some classical and state-of-the-art methods, the proposed model achieves optimal results, which reveals its effectiveness and feasibility.

Simulation-driven Bearing Fault Diagnosis for Condition Monitoring without Faulty Data

The failure of rolling element bearings in complex mechanical systems is a significant cause of mechanical failures, leading to decreased productivity and safety risks. Deep learning has shown promising results in bearing fault diagnosis, but the predictive performance depends on highquality data. Domain adaptation has been studied to solve this problem, but it still has limitations when applied to real-world industrial applications. In this study, we propose a deep learning-based domain generalization framework for bearing fault diagnosis using the bearing simulation model and adversarial data augmentation method. The proposed framework was validated on a real bearing fault dataset and showed promising results in improving diagnostic performance in cases where fault data cannot be obtained or when dealing with unlearned target domains. This approach has the potential to improve industrial maintenance systems by obtaining improved generalization performance in the absence of fault datasets.

Domain fuzzy generalization networks for semi-supervised intelligent fault diagnosis under unseen working conditions

In recent years, domain adaptation methods have made remarkable achievements in fault diagnosis under variable working conditions. However, the methods usually fail when target data are unavailable for model training. Confronting with the problem of intelligent fault diagnosis for unseen working conditions, domain generalization methods have been gradually explored. Most existing domain generalization fault diagnosis methods are supervised learning models that require multiple fully labeled source domains. Few studies have been done on semi-supervised domain generalization when only partial source domains have class labels, which is generally a practical industrial scenario because labeling industrial data is a laborious work and requires scarce domain experts. Consequently, this paper proposes a novel semi-supervised domain generalization framework, named domain fuzzy generalization networks (DFGN), for intelligent fault diagnosis under unseen working conditions. The main idea of the DFGN is to enhance the capabilities of learning domain-invariant and discriminative features by proposing domain fuzzy and metric learning strategies. First, the traditional domain discriminator outputting one-dimensional domain probability is innovatively substituted by a domain classifier that predicts the domain probabilities belonging to all the source domains. Then, the domain fuzzy strategy is established in domain-adversarial training to extract the domain-invariant features with fine-grained distribution alignment. Finally, the metric learning is embedded in feature extractor to extract the discriminative features from a class-level optimization perspective. Benefited from the extracted domain-invariant and discriminative features, the proposed DFGN model exhibits strong generalization ability that can be effectively applied to intelligent fault diagnosis under unseen working conditions. The advantages and superiority of the proposed method over state-of-the-art semi-supervised domain generalization methods are confirmed by extensive generalization experiments on two bearing datasets.

Towards generalization on real domain for single image dehazing via meta-learning

Learning-based image dehazing methods are essential to assist autonomous systems in enhancing reliability. Due to the domain gap between synthetic and real domains, the internal information learned from synthesized images is usually sub-optimal for real domains. Thus, the dehazing models trained on synthetic images often suffer from severe performance drops on real-world samples. Driven by the ability to explore internal information from a few unseen-domain samples, meta-learning is commonly adopted to address this issue via test-time training, which is hyperparameter-sensitive and time-consuming. In contrast, we present a domain generalization framework based on meta-learning to dig out representative and discriminative internal properties of real hazy domains without test-time training. To obtain representative domain-specific information, we attach two entities termed adaptation network and distance-aware aggregator to our dehazing network. The adaptation network assists in distilling domain-relevant information from a few hazy samples and caching it into a collection of features. The distance-aware aggregator strives to summarize the generated features and filter out misleading information for more representative internal properties. To enhance the discrimination of distilled internal information, we present a novel loss function called domain-relevant contrastive regularization. It encourages the similarity of the internal features generated from the same domain and the distinction of the features from diverse domains. The generated representative and discriminative features are regarded as some external variables of our dehazing network to regress a particular function for a given domain. The extensive experiments on real hazy datasets validate that our proposed method has superior generalization ability than the state-of-the-art competitors.

Achieving domain generalization for underwater object detection by domain mixup and contrastive learning

The performance of existing underwater object detection methods severely degrades when they face the domain shift caused by complicated underwater environments. Due to the limited domain diversity in collected data, deep detectors easily memorize a few seen domains, which leads to low generalization ability. There are two common ideas to improve the domain generalization performance. First, it can be inferred that the detector trained on as many domains as possible is domain-invariant. Second, their hidden features should be equivalent because the images with the same semantic content are in different domains. This paper further excavates these two ideas and proposes a domain generalization framework that learns how to generalize across domains from Domain Mixup and Contrastive Learning (DMCL). First, based on the formation of underwater images, an image in one kind of underwater environment is the linear transformation of another underwater environment. Therefore, a style transfer model, which outputs a linear transformation matrix instead of the whole image, is proposed to transform images from one source domain to another, enriching the domain diversity of the training data. Second, the Mixup operation interpolates different domains on the feature level, sampling new domains on the domain manifold. Third, a contrastive loss is selectively applied to features from different domains to force the model to learn domain-invariant features but retain the discriminative capacity. With our method, detectors will be robust to domain shift. Also, a domain generalization benchmark S-UODAC2020 for detection is set up to measure the performance of our method. Comprehensive experiments on S-UODAC2020 and two object recognition benchmarks (PACS and VLCS) demonstrate that the proposed method is able to learn domain-invariant representations and outperforms other domain generalization methods. The code is available inhttps://github.com/mousecpn/DMC-Domain-Generalization-for-Underwater-Object-Detection.git

Domain-Generalized EEG Classification With Category-Oriented Feature Decorrelation and Cross-View Consistency Learning.

Generalizing the electroencephalogram (EEG) decoding methods to unseen subjects is an important research direction for realizing practical application of brain-computer interfaces (BCIs). Since distribution shifts across subjects, the performance of most current deep neural networks for decoding EEG signals degrades when dealing with unseen subjects. Domain generalization (DG) aims to tackle this issue by learning invariant representations across subjects. To this end, we propose a novel domain-generalized EEG classification framework, named FDCL, to generalize EEG decoding through category-relevant and -irrelevant Feature Decorrelation and Cross-view invariant feature Learning. Specifically, we first devise data augmented regularization through mixing the segments of same-category features from multiple subjects, which increases the diversity of EEG data by spanning the space of subjects. Furthermore, we introduce feature decorrelation regularization to learn the weights of the augmented EEG trials to remove the dependencies between their features, so that the true mapping relationship between relevant features and corresponding labels can be better established. To further distill subject-invariant EEG feature representations, cross-view consistency learning regularization is introduced to encourage consistent predictions of category-relevant features induced from different augmented EEG views. We seamlessly integrate three complementary regularizations into a unified DG framework to jointly improve the generalizability and robustness of the model on unseen subjects. Experimental results on motor imagery (MI) based EEG datasets validate that the proposed FDCL outperforms the available state-of-the-art methods.

Health Assessment of Rotating Equipment With Unseen Conditions Using Adversarial Domain Generalization Toward Self-Supervised Regularization Learning

Traditional health assessment models work well under the assumption that the test and training samples obey a similar distribution. However, it is practically impossible to eliminate domain shifts between different tasks. Thus, most work tries to establish a data-driven approach via domain adaptation to accomplish transfer learning between different operating conditions. Sufficient target data are needed to participate in the training, which may not normally be available due to most working scenarios being unseen. An adversarial domain generalization framework with regularization learning (ADGR) is proposed for the health assessment to mine latent domains. Also, the latent domain is expanded to the unseen domain as possible. More specifically, the diversity of the sample distribution is augmented by adversarial training and the maximization of the domain discrepancy between the latent and source domains. Meanwhile, self-supervised interdomain regularization and semantical consistent regularization are proposed to mitigate the feature drift of the domain classifier and semantic divergence between source and latent domains. The case study shows that the ADGR-based health assessment approach achieves competitive prediction accuracy under unseen conditions, demonstrating its potential as a diagnostic solution.

Whitening-Net: A Generalized Network to Diagnose the Faults Among Different Machines and Conditions.

Intelligent bearing diagnostic methods are developing rapidly, but they are difficult to implement due to the lack of real industrial data. A feasible way to deal with this problem is to train a network through laboratory data to mine the causality of bearing faults. This means that the constructed network can handle domain deviations caused by the change of machines, working conditions, noise, and so on which is, however, not a simple task. In response to this problem, a new domain generalization framework-Whitening-Net-was proposed in this article. This framework first defined the homologous compound domain signal as the data basis. Subsequently, the causal loss was proposed to impose regularization constraints on the network, which enhances the network's ability to mine causality. To avoid domain-specific information from interfering with causal mining, a whitening structure was proposed to whiten the domain, prompting the network to pay more attention to the causality of the signal rather than the domain noise. The results of diagnosis and interpretation proved the ability of Whitening-Net in mining causal mechanisms, which shows that the proposed network can generalize to different machines, even if the tested working conditions and bearing types are completely different from the training domains.

Instance theory as a domain-general framework for cognitive psychology

Instance theory as a domain-general framework for cognitive psychology