Bearing fault diagnosis poses challenges such as limited feature extraction, fixed activation functions, low interpretability, and unstable diagnostic performance in graph neural networks. This necessitates the development of adaptive learning and optimization strategies. This paper introduces simulated annealing (SA)-Kolmogorov-Arnold Network (KAN)-GraphSAGE, a novel graph neural network framework for feature extraction in unsupervised intelligent bearing failure diagnosis. The framework leverages the KAN with SA optimization to enhance its capabilities. It features a specially designed KAN layer that replaces traditional multilayer perceptrons, enabling adaptive nonlinear activation functions to effectively capture complex fault feature patterns. Moreover, a KAN aggregator, utilizing a message-passing mechanism, integrates neighborhood structural information with learnable transformations to enhance feature representation. Hierarchical optimization is achieved through a combination of the SA algorithm and parameter sensitivity analysis to ensure model stability. The dual interpretability framework combines the interpretability of KAN networks with knowledge graphs to enhance semantic understanding during feature learning. An unsupervised training mechanism employing contrastive learning and feature reconstruction facilitates efficient feature extraction without extensive labeled data. Validation on Case Western Reserve University and Machinery Failure Prevention Technology bearing datasets demonstrates that SA-KAN-GraphSAGE surpasses seven mainstream approaches and two state-of-the-art methods in clustering quality, representation capability, and numerical stability metrics. This study establishes a robust technical foundation for equipment health monitoring in intelligent manufacturing.

Long-horizon egocentric memory transforms continuous first-person video and audio into a searchable record of past experiences. We demonstrate two bottlenecks in existing systems: indices built from context-poor captions are unreliable for agentic search, while retrieval ignores a question's temporal intent. To address both bottlenecks, we introduce EgoCITE (Egocentric Context-augmented Indexing and Time-aware Evidence retrieval), a long-horizon agentic memory framework for egocentric QA. EgoCITE comprises three components. EgoScheme uses local multimodal context to turn fragmentary video captions and speech transcripts into self-contained atomic memory indices. EgoIndex organizes complementary action, activity, utterance, and conversation representations into searchable multi-view memory indices at multiple granularities. EgoRetrv combines semantic search with question-conditioned temporal relevance scoring and curation of retrieved evidence. We evaluate EgoCITE on EgoLifeQA, EgoMem, and EgoR1-Bench in terms of answer accuracy and target-event retrieval alignment. EgoCITE improves accuracy over agentic memory baselines by at least 4.4–14.2

Floods are major natural disasters that cause significant destruction and loss of life, especially in small catchments lacking proper monitoring. To address this, an efficient flood forecasting system is needed for timely responses. This study focuses on the Dapoling Watershed of the Huaihe River, where we developed a flood forecasting model combining long short-term memory (LSTM), a Bayesian optimization algorithm, and error-correction (EC) methods. Moreover, model performance was evaluated using various indicators like the Nash-Sutcliffe efficiency (NSE), mean absolute error, root mean square error (RMSE), and relative errors in flood volume, peak, and timing. The results indicate that: (1) LSTM is suitable for flood forecasting in the Dapoling Watershed, although its performance declines with increasing lead time, particularly in predicting peak discharge; (2) with the incorporation of the EC method, the forecasting model exhibits significant performance improvement, as evidenced by increased NSE across all lead times; and (3) the RF-LSTM demonstrates superior applicability for flood prediction in the Dapoling Watershed, consistently outperforming other EC methods across different lead times. This framework significantly boosts forecasting accuracy and lays a solid foundation for flood early-warning systems.

This paper presents Contourformer, a real-time contour-based instance segmentation algorithm. The method is fully based on the DETR paradigm and achieves end-to-end inference through iterative and progressive mechanisms to optimize contours. To improve efficiency and accuracy, we develop two novel techniques: sub-contour decoupling mechanism and contour fine-grained distribution refinement. In the sub-contour decoupling mechanism, we propose a deformable attention-based module that adaptively selects sampling regions based on the current predicted contour, enabling more effective capturing of object boundary information. Additionally, we design a multi-stage optimization process to enhance segmentation precision by progressively refining sub-contours. The contour fine-grained distribution refinement technique aims to further improve the ability to express fine details of contours. These innovations enable Contourformer to achieve stable and precise segmentation for each instance while maintaining real-time performance. Extensive experiments demonstrate the superior performance of Contourformer on multiple benchmark datasets, including SBD, COCO, and KINS. We conduct comprehensive evaluations and comparisons with existing state-of-the-art methods, showing significant improvements in both accuracy and inference speed. This work provides a new solution for contour-based instance segmentation tasks and lays a foundation for future research, with the potential to become a strong baseline method in this field. The code is available at: https://github.com/talebolano/Contourformer.

Accurate segmentation of medical images is essential for many clinical applications and is now typically achieved by training deep learning models on large annotated datasets. However, acquiring sufficient labeled images remains challenging, as pixel-level manual annotations are highly time-consuming. To substantially reduce the manual effort, we developed a novel semi-supervised segmentation method, termed dual-decoder mutual teaching (DDMT), which incorporates a smoothed exponential moving average (sEMA) scheme and a shape consistency constraint (SCC) scheme into the classical mean teacher (MT) framework. The sEMA scheme enhances the stability of the student and teacher models during training, while the SCC scheme ensures consistent learning of shape characteristics across the two different decoders within each model. With these two innovative components, DDMT achieves promising segmentation performance when trained on limited labeled images and abundant unlabeled images. Experiments on public datasets for left atrium, pancreas, and optic disc segmentation demonstrated that DDMT consistently outperforms several state-of-the-art semi-supervised learning (SSL) methods (e.g., MT, UAMT, DTC, and MCNet) across varying proportions of labeled images. The source code is publicly available at https://github.com/wmuLei/ddmt.

Point cloud upsampling can improve the quality of the initial point cloud, significantly enhancing the performance of downstream tasks such as classification and segmentation. Existing methods mostly focus on generating the geometric details of point clouds, neglecting noise suppression. To address this, we propose a novel network based on a conditional diffusion model, incorporating the Adaptive Noise Suppression (ANS) module, which we refer to as PDANS. The ANS module assigns weights to each point and determines the removal strategy based on these weights, reducing the impact of noisy points on the sampling process. The module first selects the neighborhood set for each point in the point cloud and performs a weighted sum between the point and its neighbors. It then adjusts the removal points based on the weighted sum, effectively mitigating the bias caused by outliers. We introduce the TreeTrans (TT) module to capture more correlated feature information. This module learns the interaction between high-level and low-level features, resulting in a more comprehensive and refined feature representation. Our results on several widely used benchmark datasets demonstrate that PDANS exhibits exceptional robustness in noisy point cloud processing and outperforms current state-of-the-art(SOTA) methods in terms of performance. Code is available at https://github.com/Baty2023/PDANS.

Most of the existing class imbalanced classification methods are weak in interpretability, which is necessary for models to be convincing in some specific scenarios. In this study, we propose a dynamic broad Takagi-Sugeno-Kang (TSK) fuzzy classifier based on iterative learning on progressively expanded imbalanced dataset (DB-TSK-IL) to enhance interpretability. This method begins with a zero-order TSK fuzzy sub-classifier and constructs a TSK fuzzy classifier with a broad ensemble structure through an iterative and progressive process that repeatedly integrates the misclassified instance set of the previous broad model with the training set to train the next zero-order TSK fuzzy sub-classifier. Despite its simple structure, the proposed DB-TSK-IL provides advantageous over the comparative class-imbalanced classification methods with mostly static structures in three aspects: (1) generalization performance, (2) dynamic method with different fuzzy rules activated for different instances, and (3) significantly fewer fuzzy rules, which provide DB-TSK-IL with superior interpretability. Both the theoretical analysis and experimental results validate the efficiency of this method

Computer-assisted modeling of patient-specific 3D teeth is a clinically important technology for the development of dental diagnosis and treatment. This technology often relies on accurately segmenting the target tooth and its surrounding tissues from CBCT images. Most of the previous methods consume extensive memory for generating bounding box proposals in a detection manner, while in this paper, we propose a novel stagewise tooth instance segmentation framework from localization to segmentation. Specifically, our method follows the process of tooth centroid prediction, candidate centroid analysis, and mapping of centroids to accurately localize the ROI of individual teeth, instead of generating bounding box proposals for tooth positioning regression. To improve the segmentation quality, we propose a new loss function referred to as potential energy loss, which measures the feature similarity among voxels in a neighborhood to focus more on local information, regulating potential energy to obtain optimal segmentation. Moreover, the proposed fine segmentation network introduces a dual-branch structure and spectrum filter connections to enhance hierarchical features and anti-noise capability. Experimental results demonstrate that the proposed method surpasses state-of-the-art methods with improvements of 1.05%, 5.77%, and 16.67% on average DSC, HD95, and ASSD, respectively.

The basic goal of Automatic Check-Out (ACO) task is to accurately predict the categories and quantities of products selected by customers in the check-out images. However, there is a significant domain gap between the single-product exemplars as training data and the check-out images as testing data. To mitigate the domain gap, we propose a novel method termed as Prototype Learning for Automatic Check-Out (PLACO). In PLACO, prototype learning is designed to reach the goal in two ways. Specifically, in the prototype-based classifier learning module, to fully exploit the invariance of category prototypes, the prototypes obtained from the single-product exemplars are employed to generate classifiers for classifying the proposals of check-out image. On the other side, in prototype alignment module, prototypes for both the single-product exemplar and check-out image domains are entered simultaneously to ensure intra-category compactness and inter-category sparsity. Moreover, to further improve the performance of PLACO, we develop a discriminative re-ranking module to both adjust the predicted scores of product proposals for bringing more discriminative ability in classifier learning and provide a reasonable sorting possibility by considering the fine-grained nature. Experiments are conducted on the large-scale RPC dataset for evaluations. Our PLACO obtains the optimal results in both traditional ACO task setting and incremental task setting.

With the expansion of data scale and diversity, the issue of class imbalance has become increasingly salient. The current methods, including oversampling and under-sampling, exhibit limitations in handling complex data, leading to overfitting, loss of critical information, and insufficient interpretability. In response to these challenges, we propose a broad TSK fuzzy classifier with a simplified set of fuzzy rules (B-TSK-FC) that deals with classification tasks with class-imbalanced data. Firstly, we select and optimize fuzzy rules based on their adaptability to different complex data to simplify the fuzzy rules and therefore improve the interpretability of the TSK fuzzy sub-classifiers. Secondly, the fuzzy rules are weighted to protect the information demonstrated by minority classes, thereby improving the classification performance on class-imbalanced datasets. Finally, a novel loss function is designed to derive the weights for each TSK fuzzy sub-classifier. The experimental results on fifteen benchmark datasets demonstrate that B-TSK-FC is superior to the comparative methods from the aspects of classification performance and interpretability in the scenario of class imbalance.

Automatic Check-Out (ACO) aims to accurately predict the presence and count of each category of products in check-out images, where a major challenge is the significant domain gap between training data (single-product exemplars) and test data (check-out images). To mitigate the gap, we propose a method, termed as PSP, to perform Prototype-based classifier learning from Single-Product exemplars. In PSP, by revealing the advantages of representing category semantics, the prototype representation of each product category is firstly obtained from single-product exemplars. Based on the prototypes, it then generates categorical classifiers with a background classifier to not only recognize fine-grained product categories but also distinguish background upon product proposals derived from check-out images. To further improve the ACO accuracy, we develop discriminative re-ranking to both adjust the predicted scores of product proposals for bringing more discriminative ability in classifier learning and provide a reasonable sorting possibility by considering the fine-grained nature. Moreover, a multi-label recognition loss is also equipped for modeling co-occurrence of products in check-out images. Experiments are conducted on the large-scale RPC dataset for evaluations. Our ACO result achieves 86.69%, by 6.18% improvements over state-of-the-arts, which demonstrates the superiority of PSP. Our codes are available at https://github.com/Hao-Chen-NJUST/PSP .

In the production of the galvanized cold-rolled steel sheets used for stamping car body parts, in-situ and real-time defective detecting is crucial for quality control, in which various types of defects will inevitably occur. It is challenging to improve the accuracy of defect image classification by appropriate means to assist the manual screening process better. Defects under actual production conditions are often not prominent enough in defect characteristics, and there may be a significant similarity between different defect categories. To eliminate this weakness, we propose a data-driven faulty detection model named Steel Faulty Detection Attention Net (SFDANet) that uses images of the galvanized steel surface as input to identify whether the product is qualified and automatic classification of defect types instantaneously. This method can shorten product inspection time and improve production line efficiency automatically. In addition, the attention mechanism is utilized, enhancing the performance of SFDANet. Compared with the baseline that applied the ResNet method, SFDANet achieves a noticeable improvement in the classification accuracy of the test data. The well-trained model can successfully show an improved performance than the baseline models on the multiple types of faulty. Enhanced by SFDANet with high classification accuracy, the defect rate of products is significantly reduced, and the production speed of the production line is significantly improved.

In this paper,we tackle the long-tailed visual recognition problem from the categorical prototype perspective by proposing a prototype-based classifier learning(PCL)method.Specifically,thanks to the gen-eralization ability and robustness,categorical prototypes reveal their advantages of representing the category semantics.Coupled with their class-balance characteristic,categorical prototypes also show potential for han-dling data imbalance.In our PCL,we propose to generate the categorical classifiers based on the prototypes by performing a learnable mapping function.To further alleviate the impact of imbalance on classifier gener-ation,two kinds of classifier calibration approaches are designed from both prototype-level and example-level aspects.Extensive experiments on five benchmark datasets,including the large-scale iNaturalist,Places-LT,and ImageNet-LT,justify that the proposed PCL can outperform state-of-the-arts.Furthermore,validation experiments can demonstrate the effectiveness of tailored designs in PCL for long-tailed problems.

The practical task of Automatic Check-Out (ACO) is to accurately predict the presence and count of each product in an arbitrary product combination. Beyond the large-scale and the fine-grained nature of product categories as its main challenges, products are always continuously updated in realistic check-out scenarios, which is also required to be solved in an ACO system. Previous work in this research line almost depends on the supervisions of labor-intensive bounding boxes of products by performing a detection paradigm. While, in this paper, we propose a Self-Supervised Multi-Category Counting (S2MC2) network to leverage the point-level supervisions of products in check-out images to both lower the labeling cost and be able to return ACO predictions in a class incremental setting. Specifically, as a backbone, our S2MC2 is built upon a counting module in a class-agnostic counting fashion. Also, it consists of several crucial components including an attention module for capturing fine-grained patterns and a domain adaptation module for reducing the domain gap between single product images as training and check-out images as test. Furthermore, a self-supervised approach is utilized in S2MC2 to initialize the parameters of its backbone for better performance. By conducting comprehensive experiments on the large-scale automatic check-out dataset RPC, we demonstrate that our proposed S2MC2 achieves superior accuracy in both traditional and incremental settings of ACO tasks over the competing baselines.

A clustering-based undersampling (CUS) and distance-based near-miss method are widely used in current imbalanced learning algorithms, but this method has certain drawbacks. In particular, the CUS does not consider the influence of the distance factor on the majority of instances, and the near-miss method omits the inter-class(es) within the majority of samples. To overcome these drawbacks, this study proposes an undersampling method combining distance measurement and majority class clustering. Resampling methods are used to develop an ensemble-based imbalanced-learning algorithm called the clustering and distance-based imbalance learning model (CDEILM). This algorithm combines distance-based undersampling, feature selection, and ensemble learning. In addition, a cluster size-based resampling (CSBR) method is proposed for preserving the original distribution of the majority class, and a hybrid imbalanced learning framework is constructed by fusing various types of resampling methods. The combination of CDEILM and CSBR can be considered as a specific case of this hybrid framework. The experimental results show that the CDEILM and CSBR methods can achieve better performance than the benchmark methods, and that the hybrid model provides the best results under most circumstances. Therefore, the proposed model can be used as an alternative imbalanced learning method under specific circumstances, e.g., for providing a solution to credit evaluation problems in financial applications.

The extraction of local texture information using the traditional Local Binary Mode (LBP) is limited, and it ignores the representation of global texture information, which leads to an unsatisfactory outcome for the texture classification task. Local Binary Mode (LBP) has been widely used in texture classification. This paper utilizes LBPV to resolve this issue (Local Binary Pattern Variance) and proposes a novel adaptive weight joint multi-scale LBPV2 texture picture classification algorithm. The typical variance weight is replaced by the square of covariance as the cumulative weight of the histogram in this method, and the multi-scale texture information is retrieved using an adaptive weight and multi-scale scheme. Thus, the texture classification performance is further improved. Simulation experiments on the commonly used Outex reference texture database show that the proposed adaptive weight combined with multi-scale LBPV2 can significantly improve the performance of texture classification. In the fields of computer vision and pattern recognition, texture analysis is a fundamental visual problem with a wide range of applications, including object detection, remote sensing, content-based image retrieval, and medical picture analysis. For various research questions, numerous academics have put forth various LBP versions in recent years. The dominant LBP [2] model was proposed by Liao et al. in 2009, and it was empirically chosen as the best model out of all the models. Guo et al. proposed LBPV [3], which expresses local contrast information into the straight square of texture images using a local variance confidence and global matching scheme. In order to increase classification performance, the author also proposed a Completed Local Binary Pattern (CLBP) [4] in the same year. This pattern combines three complimentary groups—CLBPS, CLBPM, and CLBPC—using a combined probability distribution. To enhance the traditional local binary pattern's noise resistance and texture expression, Liu et al. proposed the extension of LBP [5] in 2012. Relevant scholars developed a pixel block sampling structure and local neighborhood intensity relationship model for texture classification in 2013 on the basis of conventional LBP [6], and produced notable results. More recently, in 2017, a multiscale LBP [7] was presented, transcending the constraints of conventional LBP representation and not only reflecting the microscopic texture structure but also effectively expressing the macroscopic texture structure of bigger areas. Although LBP and its variations have produced a number of remarkable texture classification results, there are still a lot of possible shortcomings. The expression of nearly entirely lost global information, for instance, results in inadequate texture classification results because most LBP versions are only capable of representing local texture information [1]. In order to significantly improve texture performance, this paper proposed a new method of adaptive weight joint multi-scale LBPV2 for classifying texture images. It did this by introducing adaptive weight classification and by replacing the variance weight in the LBPV calculation method with the variance square as the cumulative weight of the histogram.

Pathology image analysis is an essential procedure for clinical diagnosis of numerous diseases. To boost the accuracy and objectivity of the diagnosis, nowadays, an increasing number of intelligent systems are proposed. Among these methods, random field models play an indispensable role in improving the investigation performance. In this review, we present a comprehensive overview of pathology image analysis based on the Markov Random Fields (MRFs) and Conditional Random Fields (CRFs), which are two popular random field models. First of all, we introduce the framework of two random field models along with pathology images. Secondly, we summarize their analytical operation principle and optimization methods. Then, a thorough review of the recent articles based on MRFs and CRFs in the field of pathology is presented. Finally, we investigate the most commonly used methodologies from the related works and discuss the method migration in computer vision.

目的 使用条件随机场提高宫颈癌病理图像的分类准确率,用以解决传统人工诊断因受主观因素影响过大而造成的判断准确率低的问题.方法 首先,对图像进行了灰度化及网格化的预处理;其次,提取了灰度共生矩阵特征以表征宫颈癌组织病理学图像块;接着,为条件随机场的一元势和二元势选择有效的特征向量组合作为其纹理特征;最后,使用产生的势,通过设计的条件随机场模型预测最终的图像级分类结果.结果 该方法在通过免疫组化染色的宫颈癌组织病理学图像样本上获得82%的准确率.结论 显示出较好的分类准确率,说明计算机辅助诊断可以广泛应用于组织病理学图像的分类与诊断,并且该方法可以显著减轻医生负担、提高医生的工作效率以及判断的准确率.

To apply topological information to solve a cervical histopathology image clustering (CHIC) problem, a graph based unsupervised learning (GBUL) approach is proposed in this paper. First, the GBUL method applies color features and k -means clustering to carry out a first-stage “coarse” clustering. Then, a skeletonization based node generation (SBNG) approach is introduced to approximate the distribution of cervical cell nuclei. Thirdly, based on the SBNG nodes, multiple graphs are constructed. Next, graph features are extracted based on the constructed graphs. Finally, k -means clustering is used again for the second-stage clustering. In the experiment, a practical Hematoxylin–eosin staining cervical histopathology image dataset with 40 whole-slide imaging images is tested, obtaining a promising CHIC result and showing a huge potential in the cancer risk prediction field.

Cervical cancer is a very common and fatal cancer in women, but it can be prevented through early examination and treatment. Cytopathology images are often used to screen for cancer. Then, because of the possibility of artificial errors due to the large number of this method, the computer-aided diagnosis system based on deep learning is developed. The image input required by the deep learning method is usually consistent, but the size of the clinical medical image is inconsistent. The internal information is lost after resizing the image directly, so it is unreasonable. A lot of research is to directly resize the image, and the results are still robust. In order to find a reasonable explanation, 22 deep learning models are used to process images of different scales, and experiments are conducted on the SIPaKMeD dataset. The conclusion is that the deep learning method is very robust to the size changes of images. This conclusion is also validated on the Herlev dataset.