- Research Article
45
- 10.1016/j.patcog.2021.108238
Universal multi-Source domain adaptation for image classification
- Aug 14, 2021
- Pattern Recognition
- Yueming Yin + 3 more +3
Universal multi-Source domain adaptation for image classification
Abstract Modern neural networks (NNs) often do not generalize well in the presence of a "covariate shift''; that is, in situations where the training and test data distributions differ, but the conditional distribution of classification labels given the data remains unchanged. In such cases, NN generalization can be reduced to a problem of learning domain-invariant features. Domain adaptation (DA) methods include a broad range of techniques aimed at achieving this; however, these methods have struggled with the need for hyperparameter tuning, which then incurs computational costs. In this work, we introduce SIDDA, an out-of-the-box DA training algorithm built upon the Sinkhorn divergence, that can achieve effective domain alignment with minimal hyperparameter tuning and computational overhead. We demonstrate the efficacy of our method on multiple simulated and real datasets of varying complexity, including simple shapes, handwritten digits, and real astronomical observations. These datasets exhibit covariate shifts due to noise, blurring, and differences between telescopes. SIDDA is compatible with a variety of NN architectures, and it works particularly well in improving classification accuracy and model calibration when paired with equivariant NNs (ENNs). We find that SIDDA enhances the generalization capabilities of NNs, achieving up to a 40% improvement in classification accuracy on unlabeled target data. We also study the efficacy of DA on ENNs with respect to the varying orders of the dihedral group, and find that the model performance improves with higher orders. Finally, we find that SIDDA enhances model calibration on both source and target data, with the most significant gains in the unlabeled target domain---achieving over an order of magnitude improvement in the ECE and Brier score. SIDDA's versatility across various NN models and datasets, combined with its automated approach to domain alignment, has the potential to significantly advance multi-dataset studies by enabling the development of highly generalizable models
Universal multi-Source domain adaptation for image classification
Universal multi-Source domain adaptation for image classification
Unsupervised Domain Adaptation for Image Classification and Object Detection Using Guided Transfer Learning Approach and JS Divergence.
Unsupervised domain adaptation (UDA) is a transfer learning technique utilized in deep learning. UDA aims to reduce the distribution gap between labeled source and unlabeled target domains by adapting a model through fine-tuning. Typically, UDA approaches assume the same categories in both domains. The effectiveness of transfer learning depends on the degree of similarity between the domains, which determines an efficient fine-tuning strategy. Furthermore, domain-specific tasks generally perform well when the feature distributions of the domains are similar. However, utilizing a trained source model directly in the target domain may not generalize effectively due to domain shift. Domain shift can be caused by intra-class variations, camera sensor variations, background variations, and geographical changes. To address these issues, we design an efficient unsupervised domain adaptation network for image classification and object detection that can learn transferable feature representations and reduce the domain shift problem in a unified network. We propose the guided transfer learning approach to select the layers for fine-tuning the model, which enhances feature transferability and utilizes the JS-Divergence to minimize the domain discrepancy between the domains. We evaluate our proposed approaches using multiple benchmark datasets. Our domain adaptive image classification approach achieves 93.2% accuracy on the Office-31 dataset and 75.3% accuracy on the Office-Home dataset. In addition, our domain adaptive object detection approach achieves 51.1% mAP on the Foggy Cityscapes dataset and 72.7% mAP on the Indian Vehicle dataset. We conduct extensive experiments and ablation studies to demonstrate the effectiveness and efficiency of our work. Experimental results also show that our work significantly outperforms the existing methods.
Read moreToward Open-World Domain Adaptation via Iteratively Contrastive Learning and Clustering.
The open-set domain adaptation (DA) aims to address both covariate shift and category shift between a labeled source domain and an unlabeled target domain. Nevertheless, existing open-set DA methods always ignore the demand for discovering novel classes that are not present in the source domain and simply reject them as "unknown" sets without further exploration, which motivates us to understand the unknown sets more specifically. In this article, we present a more challenging open-world DA problem that recognizes seen classes while discovering novel classes in the target domain. To address this problem, we propose a novel framework that converts this problem into a clustering task via contrastive learning to learn pairwise relationships among the instances. More specifically, our method consists of two iterative steps. The semi-supervised clustering step clusters the unlabeled target data and separates it into seen and novel classes. In the contrastive learning step, based on the cluster assignments, we design tailored contrastive losses that learn pairwise relationships to reduce domain discrepancy and discover novel classes. Our method can be optimized as an example of expectation maximization (EM). We establish several baselines by extending related work. Our method obtains the superior performance on five public datasets, benchmarking this challenging setting for future research.
Read moreAdversarial Robustness for Unsupervised Domain Adaptation
Extensive Unsupervised Domain Adaptation (UDA) studies have shown great success in practice by learning transferable representations across a labeled source domain and an unlabeled target domain with deep models. However, current work focuses on improving the generalization ability of UDA models on clean examples without considering the adversarial robustness, which is crucial in real-world applications. Conventional adversarial training methods are not suitable for the adversarial robustness on the unlabeled target domain of UDA since they train models with adversarial examples generated by the supervised loss function. In this work, we propose to leverage intermediate representations learned by robust ImageNet models to improve the robustness of UDA models. Our method works by aligning the features of the UDA model with the robust features learned by ImageNet pre-trained models along with domain adaptation training. It utilizes both labeled and unlabeled domains and instills robustness without any adversarial intervention or label requirement during domain adaptation training. Our experimental results show that our method significantly improves adversarial robustness compared to the baseline while keeping clean accuracy on various UDA benchmarks.
Read moreTowards robust sensing and recognition : from statistical learning to transfer learning
Sensing and recognition technologies play an essential role in smart cities. Big data is acquired by various sensors deployed in almost every corner of our daily lives, which quantifies human activities and social systems. Smart sensing significantly facilitates people's lives and promotes the development of smart cities. Such success is due to not only the sufficient data by a variety of sensors but also the development of recognition technology driven by the progressive machine learning. As different sensors offer heterogeneous data, traditional statistical models may not be capable of tackling the high-dimensional data. To this end, neural networks with high fitting ability were proposed, which further evolved to deep neural networks as the neural layers become deeper and deeper for better capacity of feature extraction. Using deep learning, data of distinct sensors can be learned jointly for accurate recognition. \n \nNevertheless, towards robust smart sensing and recognition technology, there are still some bottlenecks that hinder the sensing performance and restrain the application scenarios. In this thesis, we propose systemic solutions and algorithms to overcome the longstanding challenges from two perspectives: sensing tools and recognition models. \n \nFrom the perspective of sensing tools, different sensors are applied to their suitable scenarios. For instance, inertial Measurement Unit (IMU) is utilized for mobile devices and cameras take charge of surveillance. However, in smart homes, existing sensors have their intrinsic shortcomings such as the privacy issue of camera, the requirement of carrying devices for IMU and low granularity of data for Pyroelectric Infrared Sensor (PIR). To obtain a cost-effective, privacy-preserving and fine-grained sensing tools, this thesis pioneers a WiFi-based sensing platform which extracts Channel State Information (CSI) data directly from routers instead of specific network cards. CSI describes the propagation situation of wireless signals which can be affected by human motions. Therefore, these patterns of variant signals render WiFi-based sensing to be possible. The proposed platform is an IoT-enabled platform which leverages off-the-shelf (COTS) WiFi routers for sensing. To demonstrate its effectiveness, we develop an occupancy detection system and a sedentary behavior monitoring system, and evaluate their performance in real-world scenarios. Traditional statistical features are employed for static and dynamic activity recognition. We further propose a novel approach for a more fine-grained recognition application --- WiFi-based gesture recognition using the platform. Since gestures are too subtle to capture, many environmental noises are recorded in CSI samples. Extracting effective features for recognition from noisy data is a non-trivial task using deep neural networks. To this end, we propose a Siamese deep model with both spatial and temporal feature extractors, which discards the intrinsic noises of CSI data during feature learning. The proposed method also allows user to fine-tune the system using few samples, and thus is user-friendly. \n \nFrom the perspective of recognition, this thesis focuses on dealing with the domain shift problem that hinders the generalization of recognition models. For instance, in WiFi-based activity recognition, the model trained in one environment cannot generalize well in a new environment due to the spatial shift existing in different environments. In visual sensing, images are captured by kinds of cameras with different brands and parameters, which hinders the performance of visual recognition models even more severely. To deal with these problems, domain adaptation minimizes the distribution discrepancy between source domain (e.g. the original environment for training) and target domain (e.g. the new environment for testing). This thesis contributes to four novel algorithms for real-world domain adaptation scenarios. \n \nFirstly, we formulate the Edge Domain Adaptation (EDA) problem that addresses the domain shift between a model trained in center server and the model running in edge devices. Then MobileDA is proposed to deal with this problem by joint minimization of domain disparity and model complexity. It bridges the gap between domain adaptation algorithms and edge computing. \n \nSecondly, current domain adaptation methods require careful manual hyper-parameter tuning, which is not realistic for unmanned systems such as smart cars. To stabilize the training of adversarial domain adaptation, Max-margin Domain-Adversarial Training (MDAT) is developed in this thesis to realize stable convergence without cumbersome manuals. \n \nThirdly, regarding the cross-domain feature learning, we further study the transferability and discriminability of deep features generated by domain adaptation models, which achieves a better performance due to the preservation of feature discriminability. This motivates us to propose Asymmetric Adversarial Domain Adaptation (AADA) that upgrades MDAT to a more generic framework. AADA also contributes to explaining the underlying mechanism behind MDAT. \n \nFourthly, Imbalanced Domain Adaptation (IDA) is proposed for the real-world imbalanced data since label shift commonly occurs to realistic data distribution. As the source domain and target domain may have imbalanced classes, IDA refers to the situation when covariate shift and label shift exist simultaneously. Then the proposed Cluster-level Discrepancy Minimization (CDM) conducts domain alignment after unsupervised clustering, which effectively aligns the domain shift and the label shift. Four algorithms for domain adaptation have been evaluated on public visual benchmark (e.g. Image-CLEF and Office-Home) and WiFi-based gesture recognition dataset. They achieve state-of-the-art performances as well as the specific goals such as training stability and better discriminant features. \n \nIn summary, the thesis contributes to robust sensing and recognition technology. For smart sensing, the WiFi-based sensing platform provides fine-grained CSI data, which can empower many applications including human activity recognition, gesture recognition, person identification, crowd counting and indoor localization. WiFI-based sensing is necessarily complementary to seamless sensing in smart homes or buildings. For robust recognition models, four domain adaptation algorithms are proposed in the thesis. They greatly give impetus to dealing with real-world domain adaptation scenarios such as long-tailed data or automatic model adaptation in edge devices. These algorithms can be further applied to improving the generalization abilities of deep models for WiFi-based sensing and visual sensing including object recognition, object detection, sentiment segmentation and person re-identification.
Read moreDomain Adaptation and Domain Generalization with Representation Learning
<p>Machine learning has achieved great successes in the area of computer vision, especially in object recognition or classification. One of the core factors of the successes is the availability of massive labeled image or video data for training, collected manually by human. Labeling source training data, however, can be expensive and time consuming. Furthermore, a large amount of labeled source data may not always guarantee traditional machine learning techniques to generalize well; there is a potential bias or mismatch in the data, i.e., the training data do not represent the target environment. To mitigate the above dataset bias/mismatch, one can consider domain adaptation: utilizing labeled training data and unlabeled target data to develop a well-performing classifier on the target environment. In some cases, however, the unlabeled target data are nonexistent, but multiple labeled sources of data exist. Such situations can be addressed by domain generalization: using multiple source training sets to produce a classifier that generalizes on the unseen target domain. Although several domain adaptation and generalization approaches have been proposed, the domain mismatch in object recognition remains a challenging, open problem – the model performance has yet reached to a satisfactory level in real world applications. The overall goal of this thesis is to progress towards solving dataset bias in visual object recognition through representation learning in the context of domain adaptation and domain generalization. Representation learning is concerned with finding proper data representations or features via learning rather than via engineering by human experts. This thesis proposes several representation learning solutions based on deep learning and kernel methods. This thesis introduces a robust-to-noise deep neural network for handwritten digit classification trained on “clean” images only, which we name Deep Hybrid Network (DHN). DHNs are based on a particular combination of sparse autoencoders and restricted Boltzmann machines. The results show that DHN performs better than the standard deep neural network in recognizing digits with Gaussian and impulse noise, block and border occlusions. This thesis proposes the Domain Adaptive Neural Network (DaNN), a neural network based domain adaptation algorithm that minimizes the classification error and the domain discrepancy between the source and target data representations. The experiments show the competitiveness of DaNN against several state-of-the-art methods on a benchmark object dataset. This thesis develops the Multi-task Autoencoder (MTAE), a domain generalization algorithm based on autoencoders trained via multi-task learning. MTAE learns to transform the original image into its analogs in multiple related domains simultaneously. The results show that the MTAE’s representations provide better classification performance than some alternative autoencoder-based models as well as the current state-of-the-art domain generalization algorithms. This thesis proposes a fast kernel-based representation learning algorithm for both domain adaptation and domain generalization, Scatter Component Analysis (SCA). SCA finds a data representation that trades between maximizing the separability of classes, minimizing the mismatch between domains, and maximizing the separability of the whole data points. The results show that SCA performs much faster than some competitive algorithms, while providing state-of-the-art accuracy in both domain adaptation and domain generalization. Finally, this thesis presents the Deep Reconstruction-Classification Network (DRCN), a deep convolutional network for domain adaptation. DRCN learns to classify labeled source data and also to reconstruct unlabeled target data via a shared encoding representation. The results show that DRCN provides competitive or better performance than the prior state-of-the-art model on several cross-domain object datasets.</p>
Read moreDomain Adaptation and Domain Generalization with Representation Learning
<p>Machine learning has achieved great successes in the area of computer vision, especially in object recognition or classification. One of the core factors of the successes is the availability of massive labeled image or video data for training, collected manually by human. Labeling source training data, however, can be expensive and time consuming. Furthermore, a large amount of labeled source data may not always guarantee traditional machine learning techniques to generalize well; there is a potential bias or mismatch in the data, i.e., the training data do not represent the target environment. To mitigate the above dataset bias/mismatch, one can consider domain adaptation: utilizing labeled training data and unlabeled target data to develop a well-performing classifier on the target environment. In some cases, however, the unlabeled target data are nonexistent, but multiple labeled sources of data exist. Such situations can be addressed by domain generalization: using multiple source training sets to produce a classifier that generalizes on the unseen target domain. Although several domain adaptation and generalization approaches have been proposed, the domain mismatch in object recognition remains a challenging, open problem – the model performance has yet reached to a satisfactory level in real world applications. The overall goal of this thesis is to progress towards solving dataset bias in visual object recognition through representation learning in the context of domain adaptation and domain generalization. Representation learning is concerned with finding proper data representations or features via learning rather than via engineering by human experts. This thesis proposes several representation learning solutions based on deep learning and kernel methods. This thesis introduces a robust-to-noise deep neural network for handwritten digit classification trained on “clean” images only, which we name Deep Hybrid Network (DHN). DHNs are based on a particular combination of sparse autoencoders and restricted Boltzmann machines. The results show that DHN performs better than the standard deep neural network in recognizing digits with Gaussian and impulse noise, block and border occlusions. This thesis proposes the Domain Adaptive Neural Network (DaNN), a neural network based domain adaptation algorithm that minimizes the classification error and the domain discrepancy between the source and target data representations. The experiments show the competitiveness of DaNN against several state-of-the-art methods on a benchmark object dataset. This thesis develops the Multi-task Autoencoder (MTAE), a domain generalization algorithm based on autoencoders trained via multi-task learning. MTAE learns to transform the original image into its analogs in multiple related domains simultaneously. The results show that the MTAE’s representations provide better classification performance than some alternative autoencoder-based models as well as the current state-of-the-art domain generalization algorithms. This thesis proposes a fast kernel-based representation learning algorithm for both domain adaptation and domain generalization, Scatter Component Analysis (SCA). SCA finds a data representation that trades between maximizing the separability of classes, minimizing the mismatch between domains, and maximizing the separability of the whole data points. The results show that SCA performs much faster than some competitive algorithms, while providing state-of-the-art accuracy in both domain adaptation and domain generalization. Finally, this thesis presents the Deep Reconstruction-Classification Network (DRCN), a deep convolutional network for domain adaptation. DRCN learns to classify labeled source data and also to reconstruct unlabeled target data via a shared encoding representation. The results show that DRCN provides competitive or better performance than the prior state-of-the-art model on several cross-domain object datasets.</p>
Read moreLabel-efficient visual recognition with transfer learning
Visual recognition, encompassing tasks such as image classification, object detection, and semantic segmentation, has been a fundamental challenge in computer vision research. It also serves as the foundation for numerous applications, including autonomous driving, remote sensing, and robotics. The emergence of deep learning has significantly advanced this field, enabling remarkable achievements through the use of end-to-end trainable deep neural networks (DNNs). However, this remarkable progress comes at the cost of requiring vast amounts of annotated training data, the collection of which is often extremely labor-intensive and time-consuming. This constraint presents a significant barrier to the scalability and adaptability of various deep learning approaches. Transfer learning, which aims to transfer the off-the-shelf learned knowledge from one or multiple related source domains to target domains, is a promising way to mitigate this constraint. In this thesis, we explore label-efficient visual recognition with various transfer learning techniques, where the objective is to learn effective visual representations for target domains without any annotations by transferring the off-the-shelf knowledge learned in different but related source domains. Specifically, we investigate three representative types of transfer learning problems, including unsupervised domain adaptation, black-box unsupervised domain adaptation and test-time prompt tuning. As one of the most typical transfer learning problems, unsupervised domain adaptation aims to adapt a visual recognition model that is trained on labeled source-domain data to perform well on a related but unlabeled target-domain data. The primary challenge is to mitigate the distribution differences between the source and target domains, which are also referred to as domain gap or domain discrepancy, such that the model can generalize well on the target domain. In this thesis, we propose three innovative unsupervised domain adaptation techniques designed to efficiently transfer knowledge from labeled source domains to unlabeled target domains across a range of visual recognition tasks. Specifically, we explore diverse approaches to align domains from multiple perspectives (e.g., data or feature space alignment), thereby reducing the domain gap and enabling to perform well on unlabeled target domain. Black-box unsupervised domain adaptation aims to learn visual recognition model for target domain with the initial predictions of target data obtained by black-box source-domain model. In another word, during the adaptation process, only the model predictions of target data are available while either source data or source models are inaccessible. Compared to conventional unsupervised domain adaptation, black-box unsupervised domain adaptation offers notable advantages in terms of data privacy and flexibility, allowing adaptation to different target models regardless of the architecture or details of the source-trained black-box model. In this thesis, we analyze the challenges in black-box domain adaptation and design a novel approach for effectively exploiting the initial predictions of target data, leading to better adapted model on unlabeled target data. We further explore label-efficient visual recognition in the recent era of vision-language models. Despite the impressive zero-shot generalization capabilities of vision-language models, they always face unneglectable gaps while applied to target datasets, i.e., a target dataset typically features unique image styles and text formats tailored to its specific tasks. To circumvent this limitation, we explore an effective transfer learning approach for vision-language models, i.e., test-time prompt tuning, where the prompts are optimized as learnable vectors during the test stage without requiring any annotations for target data. Specifically, we examine the limitation of existing test-time prompt tuning methods, and propose a novel approach that could effectively exploit the learned knowledge in vision-language models and learn prompts for target data on-the-fly. Extensive experiments conducted over various visual recognition benchmarks indicate that our proposed methods achieve superior performance over target datasets, enabling to largely mitigate the dependence of labeled data across various visual recognition tasks.
Read moreCUDA: Contradistinguisher for Unsupervised Domain Adaptation
Humans are very sophisticated in learning new information on a completely unknown domain because humans can contradistinguish, i.e., distinguish by contrasting qualities. We learn on a new unknown domain by jointly using unsupervised information directly from unknown domain and supervised information previously acquired knowledge from some other domain. Motivated by this supervised-unsupervised joint learning, we propose a simple model referred as Contradistinguisher (CTDR) for unsupervised domain adaptation whose objective is to jointly learn to contradistinguish on unlabeled target domain in a fully unsupervised manner along with prior knowledge acquired by supervised learning on an entirely different domain. Most recent works in domain adaptation rely on an indirect way of first aligning the source and target domain distributions and then learn a classifier on labeled source domain to classify target domain. This approach of indirect way of addressing the real task of unlabeled target domain classification has three main drawbacks. (i) The sub-task of obtaining a perfect alignment of the domain in itself might be impossible due to large domain shift (e.g., language domains). (ii) The use of multiple classifiers to align the distributions, unnecessarily increases the complexity of the neural networks leading to over-fitting in many cases. (iii) Due to distribution alignment, the domain specific information is lost as the domains get morphed. In this work, we propose a simple and direct approach that does not require domain alignment. We jointly learn CTDR on both source and target distribution for unsupervised domain adaptation task using contradistinguish loss for the unlabeled target domain in conjunction with supervised loss for labeled source domain. Our experiments show that avoiding domain alignment by directly addressing the task of unlabeled target domain classification using CTDR achieves state-of-the-art results on eight visual and four language benchmark domain adaptation datasets.
Read moreDeep transferable label propagation with prototypical augmentation.
Domain adaptation (DA) seeks to utilize ample labeled data from a source domain to boost the generalization capability of models on an unlabeled target domain with divergent data distributions. Label Propagation (LP) has emerged as an efficient semi-supervised learning paradigm for DA, transferring labels between the source and target domains based on a similarity graph. Nevertheless, existing LP-based DA methods still face significant challenges: (1) Semantic insufficiency in the source training domain impairs the performance of classes with sparse structures, particularly minority classes; (2) Generated pseudo-labels exhibit low reliability due to ambiguous feature distributions; (3) The two-phase architecture decouples domain-invariant feature learning from label propagation, thus failing to achieve mutual enhancement between these two processes for DA tasks; (4) Sample-level graph construction incurs prohibitive computational costs and poor scalability when handling large-scale datasets. To address these issues, we propose a novel DA strategy, Deep Transferable Label Propagation (DTLP), that integrates prototypical augmentation techniques. Specifically, DTLP embeds three core modules into a unified end-to-end system: (1) Prototype-guided feature augmentation, termed Prototypical Augmentation (ProAug), which enriches the semantic content of the source domain by interpolating samples with class prototypes to mitigate semantic deficiency; (2) Prototype graph-based label propagation, which constructs a class-level prototypical graph rather than a sample-level one to reduce computational complexity and alleviate class imbalance; (3) Domain alignment via prototypical contrastive learning, which facilitates dynamic mutual optimization between domain-invariant feature extraction and robust label propagation while narrowing domain discrepancy. Comprehensive experiments on various benchmark datasets demonstrate that the proposed DTLP outperforms state-of-the-art LP-based DA methods, validating its effectiveness and generalizability.
Read morePartially Zero-shot Domain Adaptation from Incomplete Target Data with Missing Classes
We tackle a domain adaptation problem under partially zero-shot setting. In this setting, a certain subset of classes is missing in the unlabeled target data, while all classes appear in the labeled source data, and the goal is to discriminate all classes at the target domain. To solve this problem, we utilize an adversarial training scheme and adopt instance weighting to estimate the loss related to unavailable target data in the missing classes. The instance weight is computed on the basis of the prediction of deep neural networks, implying which instance would be similar to unseen data and having useful information for the loss estimation. This estimation makes it possible to explicitly consider all classes during the domain adaptation training even in the partially zero-shot setting, which leads to accurate adaptation between domains. Experimental results with several benchmark datasets validate the advantage of our method.
Read moreSource-Free Multidomain Adaptation With Fuzzy Rule-Based Deep Neural Networks
Unsupervised domain adaptation deals with a task from an unlabeled target domain by leveraging the knowledge gained from labeled source domain(s). The fuzzy system is adopted in domain adaptation to better tackle the uncertainty caused by information scarcity in the transfer. Most existing fuzzy and nonfuzzy domain adaptation methods depend on data-level distribution matching to eliminate the domain shift. However, data sharing can trigger privacy concerns. This situation results in the unavailability of source data, wherein most domain adaptation methods cannot be applied. Source-free domain adaptation is then proposed to handle this problem. But the existing source-free domain adaptation methods rarely deal with any soft information component due to data imprecision. Besides, fewer methods handle multiple source domains that provide richer transfer information. Thus, in this article, we propose source-free multidomain adaptation with fuzzy rule-based deep neural networks, which takes advantage of a fuzzy system to handle data uncertainty in domain adaptation without source data. To learn source private models with high generality, which is important to collect low-noise pseudotarget labels, auxiliary tasks are designed by jointly training source models from multiple domains, which share source parameters and fuzzy rules while protecting source data. To transfer fuzzy rules and fit source private parameters to the target domain, self-supervised learning and anchor-based alignment are built to force target data into source feature spaces. Experiments on real-world datasets under both homogeneous and heterogeneous label space scenarios are carried out to validate the proposed method. The results indicate the superiority of the proposed fuzzy rule-based source-free multidomain adaptation method.
Read moreLeveraging topology for domain adaptive road segmentation in satellite and aerial imagery
Leveraging topology for domain adaptive road segmentation in satellite and aerial imagery
Multi-Source Collaborative Contrastive Learning for Decentralized Domain Adaptation
Unsupervised multi-source domain adaptation aims to obtain a model working well on the unlabeled target domain by reducing the domain gap between the labeled source domains and the unlabeled target domain. Considering the data privacy and storage cost, data from multiple source domains and target domain are isolated and decentralized. This data decentralization scenario brings the difficulty of domain alignment for reducing the domain gap between the decentralized source domains and target domain, respectively. For conducting domain alignment under the data decentralization scenario, we propose Multi-source Collaborative Contrastive learning for decentralized Domain Adaptation (MCC-DA). The models from other domains are used as the bridge to reduce the domain gap. On the source domains and target domain, we penalize the inconsistency of data features extracted from the source domain models and target domain model by contrastive alignment. With the collaboration of source domain models and target domain model, the domain gap between decentralized source domains and target domain is reduced without accessing the data from other domains. The experiment results on multiple benchmarks indicate that our method can reduce the domain gap effectively and outperform the state-of-the-art methods significantly.
Read moreAdaptive centroid prototype-based domain adaptation for fault diagnosis of rotating machinery without source data
Adaptive centroid prototype-based domain adaptation for fault diagnosis of rotating machinery without source data