Local Discrepancies Research Articles

ALNet: An adaptive channel attention network with local discrepancy perception for accurate indoor visual localization

Visual localization, a fundamental component of several computer vision tasks, has been predominantly realized by scene coordinate regression (SCoRe) techniques. These methods leverage neural networks for scene coordinates prediction, followed by a PnP algorithm to recover the 6-DOF camera pose. However, similar image patches are prevalent in indoor scenes, which results in the extraction of comparable features for the regression of different scene coordinates. As a result, the localization accuracy is severely declined. In this work, we develop ALNet, a novel SCoRe method that incorporates a local discrepancy perception module (LDPM) and an adaptive channel attention module (ACAM) to address this challenge. For LDPM, our key insight lies in that scene attributes around different similar image patches are inconsistent. Technically, for each image patch, LDPM identifies a certain number of the most dissimilar patches around it and computes difference vectors to enrich its own features, thereby enabling the differentiation of similar image patches. Considering geometric attributes are beneficial for distinguishing similar patches while semantic context is conducive to encoding regression issues, integrating multi-level features is an effective approach to elevate the localization accuracy. Therefore, ACAM concatenates multi-level features together and leverages both average pooling and max pooling to generate reliable channel-wise weighting coefficient, thereby modeling the correlation among channels to integrate multi-level features effectively. Comprehensive experiments are conducted on mainstream indoor localization benchmarks and an actual environment, showing that ALNet achieves impressive performance. Source code and the experimental results video are available at https://github.com/DAMMONGAO/alnet.

Few-shot condition diagnosis of rolling bearing using adversarial transfer network with class aggregation-guided

For the challenge of fault identification under limited labeled data in engineering applications, a novel adversarial transfer network with class aggregation-guided (ATN-CA) is proposed for few-shot condition diagnosis of bearings. The ATN-CA can focus on the discrepancy features of bearings by the proposed local discrepancy feature representation, which avoids that the features extracted by a single neural network may omit important fault information. Further, the proposed class aggregation-guided strategy uses the semantic information of signals to guide the dynamic adaptation of marginal and conditional distributions of source and target data, which shortens the distribution distance of the same category in different domains, thus completing the transfer diagnosis. By comparing with some existing methods on the artificial and real bearing fault datasets, results show the proposed method has the highest test precision and the smallest accuracy deviation in the transfer diagnosis of bearings.

A deep targeted transfer network with clustering pseudo-label learning for fault diagnosis across different Machines

In recent years, deep transfer models have addressed the issue of distribution shift between the source and target domains by learning domain-invariant features. However, in cross-machine fault diagnosis scenarios, existing deep learning models struggle to fit the conditional distribution of target domain samples, limiting the performance and generalization of domain adaptation models. To address these problems, we propose a deep targeted transfer network with clustering pseudo-label learning (DTTN-CPLL). DTTN-CPLL consists of three parts. First, a deep transfer network is constructed to extract cross-domain features. To reduce the intra-class distribution gap, a clustering pseudo-label learning algorithm is proposed to create subdomain labels within the target domain features. Then, we simultaneously minimize and maximize the entropy of features and the number of linearly independent vectors in the target domain to reduce the distance between subdomain features. Finally, under the constraint of local maximum mean discrepancy, we reduce the distribution discrepancy between the source and target domains at the subdomain feature level. We conducted 12 cross-machine transfer tasks on three open bearing datasets and a private high-speed train traction motor bearing dataset. The results demonstrate that, compared to other state-of-the-art models, DTTN-CPLL is effective and superior in cross-machine fault diagnosis.

PatchCL-AE: Anomaly detection for medical images using patch-wise contrastive learning-based auto-encoder

Anomaly detection is an important yet challenging task in medical image analysis. Most anomaly detection methods are based on reconstruction, but the performance of reconstruction-based methods is limited due to over-reliance on pixel-level losses. To address the limitation, we propose a patch-wise contrastive learning-based auto-encoder for medical anomaly detection. The key contribution is the patch-wise contrastive learning loss that provides supervision on local semantics to enforce semantic consistency between corresponding input–output patches. Contrastive learning pulls corresponding patch pairs closer while pushing non-corresponding ones apart between input and output, enabling the model to learn local normal features better and improve discriminability on anomalous regions. Additionally, we design an anomaly score based on local semantic discrepancies to pinpoint abnormalities by comparing feature difference rather than pixel variations. Extensive experiments on three public datasets (i.e., brain MRI, retinal OCT, and chest X-ray) achieve state-of-the-art performance, with our method achieving over 99% AUC on retinal and brain images. Both the contrastive patch-wise supervision and patch-discrepancy score provide targeted advancements to overcome the weaknesses in existing approaches.

Multi-source domain adversarial graph convolutional networks for rolling mill health states diagnosis under variable working conditions

As the rolling mill often encounters variable and complicated working conditions and shock loads, unsupervised domain adaptive (UDA) methods are imperative in its health monitoring. However, efforts of applying UDA methods on the rolling mill are negligible, and many existing approaches have constraints in domain adaptation, domain label, and data construction that prevent meaningful features from being extracted. Hence, a multi-source domain adversarial graph convolutional networks framework (MSDAGCNs) is presented to overcome these challenges and combine three essential elements to achieve cross-domain health states diagnosis under variable working conditions. First, a shared feature extract module is introduced to extract common features. Then, the features are input to a multi-source feature extract module to extract the data construction from the graphs generated by a graph construction module. Meanwhile, a multi-source domain adversarial classifier module is modeled to extract multi-source invariant features and classify them. After that, the local maximum mean discrepancy is employed to align the domain categories. Next, a task classifier module integrates the results of the multi-source classifier for reliable health state diagnosis. Results on the two cases can verify that the proposed MSDAGCNs can not only outperform other state-of-the-art methods, but also extract domain-invariant knowledge. Compared with the best-performing method, the proposed method can boost accuracy by 0.53% and 0.83% in the simplest task of the two case studies, respectively. Furthermore, the arrangement of sensors on the rolling mill is discussed to select the optimal location for collecting vibrations.

Diagnosing oral and maxillofacial diseases using deep learning

The classification and localization of odontogenic lesions from panoramic radiographs is a challenging task due to the positional biases and class imbalances of the lesions. To address these challenges, a novel neural network, DOLNet, is proposed that uses mutually influencing hierarchical attention across different image scales to jointly learn the global representation of the entire jaw and the local discrepancy between normal tissue and lesions. The proposed approach uses local attention to learn representations within a patch. From the patch-level representations, we generate inter-patch, i.e., global, attention maps to represent the positional prior of lesions in the whole image. Global attention enables the reciprocal calibration of path-level representations by considering non-local information from other patches, thereby improving the generation of whole-image-level representation. To address class imbalances, we propose an effective data augmentation technique that involves merging lesion crops with normal images, thereby synthesizing new abnormal cases for effective model training. Our approach outperforms recent studies, enhancing the classification performance by up to 42.4% and 44.2% in recall and F1 scores, respectively, and ensuring robust lesion localization with respect to lesion size variations and positional biases. Our approach further outperforms human expert clinicians in classification by 10.7 % and 10.8 % in recall and F1 score, respectively.

Adversarial subdomain adaptation method based on multi-scale features for bearing fault diagnosis

Due to the variable working environment of bearings, the collected data often follow different probability distributions. It is hard to directly use the trained models to identify the bearing fault with different operating conditions. In addition, it is a high cost to label the samples for every work condition. To solve these problems, a multi-scale adversarial subdomain adaptation bearing fault diagnosis method is proposed, which is based on Continuous Wavelet Transform (CWT) and our constructed Multi-scale Adversarial SubDomain Adaptation Network (MASDAN). Firstly, to extract the features of non-stationary signals with different frequency bands, CWT is used to convert continuous vibration signals into two-dimensional time-frequency images. Secondly, to enhance the correlation of the features across frequency bands, a multi-scale ConvNeXt is proposed, which adds a multi-scale module based on ConvNeXt to obtain features with different scales. Finally, to reduce the distribution discrepancy between the source domain and the target domain and to avoid feature confusion in different domains, a domain adaptive alignment network introducing domain information is constructed. Two modules are included: the Domain Alignment Classification Network Module (DACNM) based on Multi-kernel Local Maximum Mean Discrepancy (MK-LMMD) and the Domain Adversarial Network Module (DANM) based on domain discrimination. Thus, the MASDAN consists of the multi-scale ConvNeXt module, DACNM, and DANM, which can realize the multi-scale feature extraction and adaptively align the fault features at different domains. The experimental results on the Qingdao University of Technology (QUT) bearing dataset and the Case Western Reserve University (CWRU) bearing dataset demonstrate the proposed method can effectively diagnose bearing faults under different operating conditions.

FESNet: Spotting Facial Expressions Using Local Spatial Discrepancy and Multi-Scale Temporal Aggregation

FESNet: Spotting Facial Expressions Using Local Spatial Discrepancy and Multi-Scale Temporal Aggregation

Unsupervised rolling bearing fault diagnosis method across working conditions based on multiscale convolutional neural network

In practical engineering, the features of rolling bearing vibration signals often vary in distribution under different working conditions, and obtaining sample labels for target working conditions can be challenging. Therefore, a multi-scale depth subdomain adaptive convolutional neural network (MS-DSACNN) fault diagnosis method is proposed. The MS-DSACNN method is based on a multi-scale feature extraction method combined with an improved 1D-ConvNeXt architecture, which fuses low- and high-level features into multi-scale fault features. A channel attention mechanism is also introduced to adaptively assign weights to the fused multi-scale features. The local maximum mean discrepancy is constructed by extracting features and their corresponding source and pseudo-label inputs for the source and target working conditions, respectively, to construct the loss function for backpropagation. The MS-DSACNN model was evaluated in experiments with two different loads and one variable speed dataset. The experimental results indicate that MS-DSACNN has higher feature extraction capability and diagnostic accuracy than other unsupervised cross-condition fault diagnosis techniques. Additionally, the scenario where the signal captured by the sensor contains a considerable amount of noise in actual working conditions is considered.

A progressive multi-source domain adaptation method for bearing fault diagnosis

Based on massive samples collected from various working conditions, multi-source domain adaptation-based fault diagnosis methods have been a promising way to improve the generalization capability of diagnostic models. However, insufficient multi-domain alignment and category confusion issues lead to unsatisfactory diagnostic performance in practical cross-domain scenarios. This article proposes a progressive multi-source domain adaptation method, which learns domain-uninformative and category-informative features. First, all domains are divided into multiple source-source and source-target domain pairs on different feature spaces to explore rich domain-invariant features, where local maximum mean discrepancy is introduced for domain pairwise alignment. Second, category correlation is reduced to allow ambiguous samples to obtain clear category boundaries. Extensive experimental results show the outstanding performance of our method.

Identifying and quantifying local uncertainty and discrepancy in the comparison of global cropland extent through a synergistic approach

Spatiotemporally consistent information on global cropland extent is essential for resource management and scientific research. Multiple cropland datasets derived from remotely sensed products are currently available. However, significant discrepancies and uncertainties among them lead to acreage estimates that diverge considerably from official statistics, thus constraining their applicability. To this end, a new stratified optimization synergistic approach (SOSA) is here proposed to create hybrid cropland maps of China, circa 2000–2020, by fusing five existing land cover maps (i.e., CLCD, GLC_FCS30, Globeland30, GlobalCrop, and ESA_CCI), and sub-national statistics. Given the underlying challenges associated with cost-effective large-scale cropland mapping, this approach seeks to strike a balance between data value, veracity, and affordability. SOSA streamlines the commonly used protocol procedure for determining the optimal agreement level and the best product combination. Preliminary validation of the resulting cropland maps was performed, and the evaluation demonstrated that the synergy cropland map exhibited greater spatial accuracy and closer agreement with statistics compared to any individual input map. This hints that synergistic approaches can bolster cropland mapping performance and amplify consistency with statistical data. Our results are expected to serve as a valuable reference for data users, aid in the future improvement of cropland mapping to support forward-thinking applications, and enhance our understanding and modeling of agricultural systems worldwide.

Aero-Engine Remaining Useful Life Prediction Based on Bi-Discrepancy Network

Most unsupervised domain adaptation (UDA) methods align feature distributions across different domains through adversarial learning. However, many of them require introducing an auxiliary domain alignment model, which incurs additional computational costs. In addition, they generally focus on the global distribution alignment and ignore the fine-grained domain discrepancy, so target samples with significant domain shifts cannot be detected or processed for specific tasks. To solve these problems, a bi-discrepancy network is proposed for the cross-domain prediction task. Firstly, target samples with significant domain shifts are detected by maximizing the discrepancy between the outputs of the dual regressor. Secondly, the adversarial training mechanism is adopted between the feature generator and the dual regressor for global domain adaptation. Finally, the local maximum mean discrepancy is used to locally align the fine-grained features of different degradation stages. In 12 cross@-domain prediction tasks generated on the C-MAPSS dataset, the root-mean-square error (RMSE) was reduced by 77.24%, 61.72%, 38.97%, and 3.35% on average, compared with the four mainstream UDA methods, which proved the effectiveness of the proposed method.

Comprehensive learning and adaptive teaching: Distilling multi-modal knowledge for pathological glioma grading

The fusion of multi-modal data, e.g., pathology slides and genomic profiles, can provide complementary information and benefit glioma grading. However, genomic profiles are difficult to obtain due to the high costs and technical challenges, thus limiting the clinical applications of multi-modal diagnosis. In this work, we investigate the realistic problem where paired pathology-genomic data are available during training, while only pathology slides are accessible for inference. To solve this problem, a comprehensive learning and adaptive teaching framework is proposed to improve the performance of pathological grading models by transferring the privileged knowledge from the multi-modal teacher to the pathology student. For comprehensive learning of the multi-modal teacher, we propose a novel Saliency-Aware Masking (SA-Mask) strategy to explore richer disease-related features from both modalities by masking the most salient features. For adaptive teaching of the pathology student, we first devise a Local Topology Preserving and Discrepancy Eliminating Contrastive Distillation (TDC-Distill) module to align the feature distributions of the teacher and student models. Furthermore, considering the multi-modal teacher may include incorrect information, we propose a Gradient-guided Knowledge Refinement (GK-Refine) module that builds a knowledge bank and adaptively absorbs the reliable knowledge according to their agreement in the gradient space. Experiments on the TCGA GBM-LGG dataset show that our proposed distillation framework improves the pathological glioma grading and outperforms other KD methods. Notably, with the sole pathology slides, our method achieves comparable performance with existing multi-modal methods. The code is available at https://github.com/CUHK-AIM-Group/MultiModal-learning.

Comparison of Publicly Available Bathymetric Data with Real Measurements in the Southeastern Black Sea

Being able to access accurate and reliable depth information has uncountable benefits for not only fields of oceanography, geophysics, geology, natural resources but also navigation & logistics. There is an ever-increasing demand for high-resolution bathymetric data for those fields since only a small portion of the world seas and oceans have been explored, observed, and charted so far. There are some sources which provide publicly available global bathymetric data to its stakeholders and users such as widely used European Marine Observation and Data Network (EMODnet) and General Bathymetric Chart of the Oceans (GEBCO). In general, it is challenging for researchers to select the best fit for their studies among available datasets, due to the fact that their sole reliability is not well-assessed. The purpose of our study, thus, is to compare those publicly available bathymetric data with field measurements obtained from the surveys carried out by the Office of Navigation, Hydrography and Oceanography for an area of interest in the southeastern Black Sea comprising various characteristics as to the bottom topography (i.e. homogeneous elevation, steep slope, mild slope, etc.). Validation is conducted focusing on these distinct features by means of visual assessment and quantitative comparison. Results reveal that, even though there is an overall agreement, local discrepancies are also present. Nonetheless, GEBCO and EMODnet datasets are proved to be great assets for any hydrospatial application that does not necessarily require high spatial resolution.

Rolling bearing fault diagnosis under time-varying speeds based on time-characteristic order spectrum and multi-scale domain adaptation network

In practical industrial applications, rolling bearing generally operates under variable conditions and its vibration signal significantly fluctuates in amplitude and frequency. This increases the feature distribution differences of the bearing fault samples and makes the health status identification of the bearing more difficult. To this end, a new intelligent fault diagnosis method for bearing under time-varying speed conditions is proposed based on time-characteristic order (TCO) spectrum and multi-scale domain adaptation network (MSDAN). Firstly, by using the good noise robustness and high time–frequency aggregation properties of the synchrosqueezed wave packet transform (SSWPT), the TCO spectrum method based on SSWPT is proposed to eliminate the impact of speed fluctuation, reducing the distribution shift of bearing data under time-varying speeds. Secondly, an MSDAN model based on global-local feature fusion is established to extract the domain-invariant features closely related to the bearing fault state from the TCO spectrum. Finally, the local maximum distribution discrepancy is introduced to capture the discriminative fine-grained features. The feasibility of the proposed method is verified in various transfer tasks on two different bearing datasets with time-varying speeds. Compared with some state-of-art methods, the proposed method can eliminate the sample distribution differences under time-varying speeds, significantly improving its accuracy and generalization performance in cross-domain fault diagnosis of rolling bearing.

Multi-Source Video Domain Adaptation With Temporal Attentive Moment Alignment Network

Multi-Source Domain Adaptation (MSDA) is a more practical domain adaptation scenario in real-world scenarios, which relaxes the assumption in conventional Unsupervised Domain Adaptation (UDA) that source data are sampled from a single domain and match a uniform data distribution. The MSDA is more challenging due to the existence of different domain shifts between distinct domain pairs. When considering videos, the negative transfer would be provoked by spatial-temporal features and can be formulated into a more challenging Multi-Source Video Domain Adaptation (MSVDA) problem. In this paper, we address the MSVDA problem by proposing a novel Temporal Attentive Moment Alignment Network (TAMAN) which aims for effective feature transfer by dynamically aligning both spatial and temporal feature moments. The TAMAN further constructs robust global temporal features by attending to dominant domain-invariant local temporal features with high local classification confidence and low disparity between global and local feature discrepancies. To facilitate future research on the MSVDA problem, we introduce comprehensive benchmarks, covering extensive MSVDA scenarios. Empirical results demonstrate a superior performance of the proposed TAMAN across multiple MSVDA benchmarks.

Learning Smooth Representation for Unsupervised Domain Adaptation.

Typical adversarial-training-based unsupervised domain adaptation (UDA) methods are vulnerable when the source and target datasets are highly complex or exhibit a large discrepancy between their data distributions. Recently, several Lipschitz-constraint-based methods have been explored. The satisfaction of Lipschitz continuity guarantees a remarkable performance on a target domain. However, they lack a mathematical analysis of why a Lipschitz constraint is beneficial to UDA and usually perform poorly on large-scale datasets. In this article, we take the principle of utilizing a Lipschitz constraint further by discussing how it affects the error bound of UDA. A connection between them is built, and an illustration of how Lipschitzness reduces the error bound is presented. A local smooth discrepancy is defined to measure the Lipschitzness of a target distribution in a pointwise way. When constructing a deep end-to-end model, to ensure the effectiveness and stability of UDA, three critical factors are considered in our proposed optimization strategy, i.e., the sample amount of a target domain, dimension, and batchsize of samples. Experimental results demonstrate that our model performs well on several standard benchmarks. Our ablation study shows that the sample amount of a target domain, the dimension, and batchsize of samples, indeed, greatly impact Lipschitz-constraint-based methods' ability to handle large-scale datasets. Code is available at https://github.com/CuthbertCai/SRDA.

Fault transfer diagnosis of rolling bearings across multiple working conditions via subdomain adaptation and improved vision transformer network

Due to often working in the environment of variable speeds and loads, it is an enormous challenge to achieve high-accuracy fault diagnosis (FD) of rolling bearings (RB) via existing approaches. In the article, a novel FD approach of RB, named IVTN-SA, is proposed by integrating subdomain adaptation (SA) and an improved vision transformer network (IVTN). To begin with, a local maximum mean discrepancy is introduced to replace the popular distribution alignment strategy of the same fault type in different domains based on adversarial learning mechanism and global maximum mean discrepancy. Then, the traditional vision transformer net is improved by employing a deformable convolution (DC) module to replace plain counterparts in existing CNN architectures and using a recurrent neural network to obtain the position encoding adaptively. The proposed method makes full use of the strong ability of SA in domain adaptation, the distinctive advantage of DC on feature extraction based on local information and the excellent performance of vision transformer in representing complicated relationships based on global information, thus realizing the fusion of local and global information and overcoming the distribution difference caused by working condition fluctuation. Two experimental cases have been conducted to verify its effectiveness in various working conditions, and the results demonstrate our proposed approach can achieve more excellent performance on diagnosis accuracy and model complexity compared with existing methods.

A Novel Periodic Cyclic Sparse Network With Entire Domain Adaptation for Deep Transfer Fault Diagnosis of Rolling Bearing

In recent years, rolling bearing fault diagnosis technique based on deep learning provides a more intelligent and reliable way for the safe operation of mechanical system. However, this technique still exists the problems of high model complexity and poor generalization ability in application. To solve the above problem, a novel periodic cyclic sparse network with entire domain adaptation (PcsNet-EDA) for deep transfer fault diagnosis of rolling bearing is proposed in this paper. The proposed periodic cyclic sparse design pattern makes the weight matrices of the convolutional layer and the fully-connected layer contain a large number of zero-weight parameters with regular arrangement, which effectively reduces the complexity of the model. The proposed EDA simultaneously considers the alignment of both global and local discrepancy, which further enhances the generalization ability of the model. In experimental validation, this paper firstly analyzes the basic diagnostic performance of PcsNet. Then, the transfer diagnostic performance of PcsNet within single-source domain scenarios based on EDA is explored, including the influence of different pre-defined sparse structures on the transfer diagnostic accuracy. The validation shows that the proposed PcsNet and the corresponding transfer model PcsNet-EDA can achieve satisfactory results on different bearing fault datasets for fault diagnosis.

Research on a Rolling Bearing Fault Diagnosis Method Based on Multi-Source Deep Sub-Domain Adaptation

Rolling bearings are the core component of rotating machinery. In order to solve the problem that the distribution of collected rolling bearing data is inconsistent during the operation of bearings under complex working conditions, which results in poor fault identification effects, a fault diagnosis method based on multi-source deep sub-domain adaptation (MSDSA) is proposed in this paper. The proposed method uses CMOR wavelet transform to transform the collected vibration signal into time–frequency maps and construct multiple sets of source–target domain data pairs, and a rolling bearing fault diagnosis network based on a multi-source deep sub-domain adaptive network is established. The network uses shared and domain-specific feature extraction networks to extract data features together. At the same time, the local maximum mean discrepancy (LMMD) was introduced to effectively capture the fine-grained information of the category. Each set of data was used to train the corresponding classifier. Finally, multiple sets of classifiers were combined to reduce the classification loss of the target domain samples at the classification boundary to achieve fault identification. In order to make the training process more stable, the network used the Ranger optimizer for parameter tuning. This paper verifies the effectiveness of the proposed method through two sets of comparative experiments. The proposed method achieves 97.78%, 99.65%, and 99.34% in three migration tasks. The experimental results show that the proposed method has a high recognition rate and generalization performance.