Search Results (95)

Occlusion remains a major challenge in person re-identification, as it often leads to incomplete or misleading visual cues. To address this issue, we propose a dual-branch occlusion-aware network (DOAN), which explicitly and implicitly enhances the model’s capability to perceive and handle occlusions. The proposed DOAN framework comprises two synergistic branches. In the first branch, we introduce an Occlusion-Aware Semantic Attention (OASA) module to extract semantic part features, incorporating a parallel channel and spatial attention (PCSA) block to precisely distinguish between pedestrian body regions and occlusion noise. We also generate occlusion-aware parsing labels by combining external human parsing annotations with occluder masks, providing structural supervision to guide the model in focusing on visible regions. In the second branch, we develop an occlusion-aware recovery (OAR) module that reconstructs occluded pedestrians to their original, unoccluded form, enabling the model to recover missing semantic information and enhance occlusion robustness. Extensive experiments on occluded, partial, and holistic benchmark datasets demonstrate that DOAN consistently outperforms existing state-of-the-art methods. Full article

Accurate diagnosis of knee joint injuries from magnetic resonance (MR) images is critical for patient care. Background/Objectives: While deep learning has advanced 3D MR image analysis, its reliance on extensive labeled datasets is a major hurdle for diverse knee pathologies. Few-shot learning (FSL) addresses this by enabling models to classify new conditions from minimal annotated examples, often leveraging knowledge from related tasks. However, creating robust 3D FSL frameworks for varied knee injuries remains challenging. Methods: We introduce MedNet-FS, a 3D FSL framework that effectively classifies knee injuries by utilizing domain-specific pre-trained weights and generalized end-to-end (GE2E) loss for discriminative embeddings. Results: MedNet-FS, with knee-MRI-specific pre-training, significantly outperformed models using generic or other medical pre-trained weights and approached supervised learning performance on internal datasets with limited samples (e.g., achieving an area under the curve (AUC) of 0.76 for ACL tear classification with k = 40 support samples on the MRNet dataset). External validation on the KneeMRI dataset revealed challenges in classifying partially torn ACL (AUC up to 0.58) but demonstrated promising performance for distinguishing intact versus fully ruptured ACLs (AUC 0.62 with k = 40). Conclusions: These findings demonstrate that tailored FSL strategies can substantially reduce data dependency in developing specialized medical imaging tools. This approach fosters rapid AI tool development for knee injuries and offers a scalable solution for data scarcity in other medical imaging domains, potentially democratizing AI-assisted diagnostics, particularly for rare conditions or in resource-limited settings. Full article

Using graphs as a data structure is a simple way to represent relationships between objects. Consequently, it has raised the need for algorithms to process, analyze, and extract meaningful information from graphs. Therefore, frequent subgraph mining (FSM) algorithms have been reported in the literature to discover interesting, unexpected, and useful patterns in graph databases. Frequent subgraph mining involves discovering subgraphs that appear no less than a user-specified threshold; this can be performed exactly or approximately. Although several algorithms for mining frequent approximate subgraphs exist, mining this type of subgraph in graph collections has scarcely been addressed. Thus, we propose AGCM-SLV, an algorithm for mining frequent approximate subgraphs within a graph collection that allows structural and label variations. Unlike other FSM approaches, our proposed algorithm tracks subgraph occurrences and their structural dissimilarities, allowing user-defined partial similarities between node and edge labels, and captures frequent approximate subgraphs (patterns) that would otherwise be overlooked. Experiments on real-world datasets demonstrate that our algorithm identifies more patterns than the most similar state-of-the-art algorithm with a shorter runtime. We also present experiments in which we add white noise to the graph collection at different levels, revealing that over 99% of the patterns extracted without noise are preserved under noisy conditions, making the proposed algorithm noise-tolerant. Full article

State estimation in distribution power systems is increasingly challenged by the proliferation of distributed energy resources (DERs), bidirectional power flows, and the growing complexity of unbalanced network topologies. Physics-Informed Neural Networks (PINNs) offer a compelling solution by integrating machine learning with the physical laws that govern power system behavior. This paper introduces a PINN-based framework for state estimation in unbalanced distribution systems, leveraging available data and embedded physical knowledge to improve accuracy, computational efficiency, and robustness across diverse operating scenarios. The proposed method is evaluated on four IEEE test feeders—IEEE 13, 34, 37, and 123—using synthetic datasets generated via OpenDSS to emulate realistic operating scenarios, and demonstrates significant improvements over baseline models. Notably, the PINN achieves up to a 97% reduction in current estimation errors while maintaining high voltage prediction accuracy. Extensive simulations further assess model performance under noisy inputs and partial observability, where the PINN consistently outperforms conventional data-driven approaches. These results highlight the method’s ability to generalize under uncertainty, accelerate convergence, and preserve physical consistency in simulated real-world conditions without requiring large volumes of labeled training data. Full article

The accurate classification of cocoa pod ripeness is critical for optimizing harvest timing, improving post-harvest processing, and ensuring consistent quality in chocolate production. Traditional ripeness assessment methods are often subjective, labor-intensive, or destructive, highlighting the need for automated, non-invasive solutions. This study evaluates the performance of R-CNN-based deep learning models—Faster R-CNN and Mask R-CNN—for the detection and segmentation of cocoa pods across four ripening stages (0–2 months, 2–4 months, 4–6 months, and >6 months) using the RipSetCocoaCNCH12 dataset, which is publicly accessible, comprising 4116 labeled images collected under real-world field conditions, in the context of precision agriculture. Initial experiments using pretrained weights and standard configurations on a custom COCO-format dataset yielded promising baseline results. Faster R-CNN achieved a mean average precision (mAP) of 64.15%, while Mask R-CNN reached 60.81%, with the highest per-class precision in mature pods (C4) but weaker detection in early stages (C1). To improve model robustness, the dataset was subsequently augmented and balanced, followed by targeted hyperparameter optimization for both architectures. The refined models were then benchmarked against state-of-the-art YOLOv8 networks (YOLOv8x and YOLOv8l-seg). Results showed that YOLOv8x achieved the highest mAP of 86.36%, outperforming YOLOv8l-seg (83.85%), Mask R-CNN (73.20%), and Faster R-CNN (67.75%) in overall detection accuracy. However, the R-CNN models offered valuable instance-level segmentation insights, particularly in complex backgrounds. Furthermore, a qualitative evaluation using confidence heatmaps and error analysis revealed that R-CNN architectures occasionally missed small or partially occluded pods. These findings highlight the complementary strengths of region-based and real-time detectors in precision agriculture and emphasize the need for class-specific enhancements and interpretability tools in real-world deployments. Full article

Multi-dimensional classification (MDC), in which the training data are concurrently associated with numerous label variables across many dimensions, has garnered significant interest recently. Most of the current MDC methods are based on the framework of supervised learning, which induces a predictive model from a large amount of precisely labeled data. So, they are challenged to obtain satisfactory learning results in the situation where the training data are not annotated with precise labels but assigned with ambiguous labels. Besides, the current MDC algorithms only consider the scenario of centralized learning, where all training data are handled at a single node for the purpose of classifier induction. However, in some real applications, the training data are not consolidated at a single fusion center, but rather are dispersedly distributed among multiple nodes. In this study, we focus on the problem of decentralized classification involving partial multi-dimensional data that have partially accessible candidate labels, and develop a distributed method called dPL-MDC for learning with these partial labels. In this algorithm, we conduct one-vs.-one decomposition on the originally heterogeneous multi-dimensional output space, such that the problem of partial MDC can be transformed into the issue of distributed partial multi-label learning. Then, by using several shared anchor data to characterize the global distribution of label variables, we propose a novel distributed approach to learn the label confidence of the training data. Under the supervision of recovered credible labels, the classifier can be induced by exploiting the high-order label dependencies from a common low-dimensional subspace. Experiments performed on various datasets indicate that our proposed method is capable of achieving learning performance in distributed partial MDC. Full article

Deep learning-based models have achieved considerable success in partial discharge (PD) fault diagnosis for power systems, enhancing grid asset safety and improving reliability. However, traditional approaches often rely on centralized training, which demands significant resources and fails to account for the impact of noisy operating conditions on Intelligent Electronic Devices (IEDs). In a gas-insulated switchgear (GIS), PD measurement data collected in noisy environments exhibit diverse feature distributions and a wide range of class representations, posing significant challenges for trained models under complex conditions. To address these challenges, we propose a Self-Supervised Asynchronous Federated Learning (SSAFL) approach for PD diagnosis in noisy IED environments. The proposed technique integrates asynchronous federated learning with self-supervised learning, enabling IEDs to learn robust pattern representations while preserving local data privacy and mitigating the effects of resource heterogeneity among IEDs. Experimental results demonstrate that the proposed SSAFL framework achieves overall accuracies of 98% and 95% on the training and testing datasets, respectively. Additionally, for the floating class in IED 1, SSAFL improves the F1-score by 5% compared to Self-Supervised Federated Learning (SSFL). These results indicate that the proposed SSAFL method offers greater adaptability to real-world scenarios. In particular, it effectively addresses the scarcity of labeled data, ensures data privacy, and efficiently utilizes heterogeneous local resources. Full article

In an increasingly interconnected world, securing large networks against cyber-threats has become paramount as cyberattacks become more rampant, difficult, and expensive to remedy. This research explores data-driven security by applying semi-supervised machine learning techniques for intrusion detection in large-scale network environments. Novel methods (including decision tree with entropy-based uncertainty sampling, logistic regression with self-training, and co-training with random forest) are proposed to perform intrusion detection with limited labeled data. These methods leverage both available labeled data and abundant unlabeled data. Extensive experiments on the CIC-DDoS2019 dataset show promising results; both the decision tree with entropy-based uncertainty sampling and the co-training with random forest models achieve 99% accuracy. Furthermore, the UNSW-NB15 dataset is introduced to conduct a comparative analysis between base models (random forest, decision tree, and logistic regression) when using only labeled data and the proposed models when using partially labeled data. The proposed methods demonstrate superior results when using 1%, 10%, and 50% labeled data, highlighting their effectiveness and potential for improving intrusion detection systems in scenarios with limited labeled data. Full article

Chemical oxygen demand (COD) serves as a key indicator of organic pollution in water bodies, and its rapid and accurate detection is crucial for environmental protection. Recently, ultraviolet–visible (UV–Vis) spectroscopy has gained popularity for COD detection due to its convenience and the absence of chemical reagents. Meanwhile, deep learning has emerged as an effective approach for automatically extracting spectral features and predicting COD. This paper proposes transforming one-dimensional spectra into two-dimensional spectrum images and employing convolutional neural networks (CNNs) to extract features and model automatically. However, training such deep learning models requires a vast dataset of water samples, alongside the complex task of labeling this data. To address these challenges, we introduce a transfer learning model based on VGG-16 for spectrum images. In this approach, parameters in the initial layers of the model are frozen, while those in the later layers are fine-tuned with the spectrum images. The effectiveness of this method is demonstrated through experiments conducted on our dataset, where the results indicate that it significantly enhances the accuracy of COD prediction compared to traditional methods and other deep learning methods such as partial least squares regression (PLSR), support vector machine (SVM), artificial neural network (ANN), and CNN-based methods. Full article

Background: The accurate identification of malignant lymph nodes in cervical ultrasound images is crucial for early diagnosis and treatment planning. Traditional evaluation metrics, such as accuracy and the Dice Similarity Coefficient (DSC), often fail to provide a realistic assessment of segmentation performance, as they do not account for partial overlaps between predictions and ground truth. This study addresses this gap by introducing the Intersection over Union (IoU) as an additional metric to offer a more comprehensive evaluation of model performance. Specifically, we aimed to develop a convolutional neural network (CNN) capable of detecting suspicious malignant lymph nodes and assess its effectiveness using both conventional and IoU-based performance metrics. Methods: A dataset consisting of 992 malignant lymph node images was extracted from 166 cervical ultrasound scans and labeled using the ImgLab annotation tool. A CNN was developed using Python, Keras, and TensorFlow and employed within the Jupyter Notebook environment. The network architecture consists of four neural layers trained to distinguish malignant lymph nodes. Results: The CNN achieved a training accuracy of 97% and a validation accuracy of 99%. The DSC score was 0.984, indicating a strong segmentation performance, although it was limited to detecting malignant lymph nodes in positive cases. An IoU evaluation applied to the test images revealed an average overlap of 74% between the ground-truth labels and model predictions, offering a more nuanced measure of the segmentation accuracy. Conclusions: The CNN demonstrated high accuracy and DSC scores, confirming its effectiveness in identifying malignant lymph nodes. However, the IoU values, while lower than conventional accuracy metrics, provided a more realistic evaluation of the model’s performance, highlighting areas for potential improvement in segmentation accuracy. This study underscores the importance of using IoU alongside traditional metrics to obtain a more reliable assessment of deep learning-based medical image analysis models. Full article

Distributed learning (DL), in which multiple nodes in an inner-connected network collaboratively induce a predictive model using their local data and some information communicated across neighboring nodes, has received significant research interest in recent years. Yet, it is challenging to achieve excellent performance in scenarios when training data instances have incomplete features and ambiguous labels. In such cases, it is essential to develop an efficient method to jointly perform the tasks of missing feature imputation and credible label recovery. Considering this, in this article, a distributed partial label missing data classification (dPMDC) algorithm is proposed. In the proposed algorithm, an integrated framework is formulated, which takes the ideas of both generative and discriminative learning into account. Firstly, by exploiting the weakly supervised information of ambiguous labels, a distributed probabilistic information-theoretic imputation method is designed to distributively fill in the missing features. Secondly, based on the imputed feature vectors, the classifier modeled by the random feature map of the ${\chi }^2$ kernel function can be learned. Two iterative steps constitute the dPMDC algorithm, which can be used to handle dispersed, distributed data with partially missing features and ambiguous labels. Experiments on several datasets show the superiority of the suggested algorithm from many viewpoints. Full article

In machine learning, when the labeled portion of data needs to be processed, a semi-supervised learning algorithm is used. A dataset with missing attribute values or labels is referred to as an incomplete information system. Addressing incomplete information within a system poses a significant challenge, which can be effectively tackled through the application of rough set theory (R-theory). However, R-theory has its limits: It fails to consider the frequency of an attribute value and then cannot the distribution of attribute values appropriately. If we consider partially labeled data and replace a missing attribute value with the multiset of all possible attribute values under the same attribute, this results in the emergence of partially labeled multiset-valued data. In a semi-supervised learning algorithm, in order to save time and costs, a large number of redundant features need to be deleted. This study proposes semi-supervised attribute selection algorithms for partially labeled multiset-valued data. Initially, a partially labeled multiset-valued decision information system (p-MSVDIS) is partitioned into two distinct systems: a labeled multiset-valued decision information system (l-MSVDIS) and an unlabeled multiset-valued decision information system (u-MSVDIS). Subsequently, using the indistinguishable relation, distinguishable relation, and dependence function, two types of attribute subset importance in a p-MSVDIS are defined: the weighted sum of l-MSVDIS and u-MSVDIS determined by the missing rate of labels, which can be considered an uncertainty measurement (UM) of a p-MSVDIS. Next, two adaptive semi-supervised attribute selection algorithms for a p-MSVDIS are introduced, which leverage the degrees of importance, allowing for automatic adaptation to diverse missing rates. Finally, experiments and statistical analyses are conducted on 11 datasets. The outcome indicates that the proposed algorithms demonstrate advantages over certain algorithms. Full article

In industrial scenarios, image segmentation is essential for accurately identifying defect regions. Recently, the emergence of foundation models driven by powerful computational resources and large-scale training data has brought about a paradigm shift in deep learning-based image segmentation. The Segment Anything Model (SAM) has shown exceptional performance across various downstream tasks, owing to its vast semantic knowledge and strong generalization capabilities. However, the feature distribution discrepancy, reliance on manually labeled prompts, and limited category information of SAM reduce its scalability in industrial settings. To address these issues, we propose PA-SAM, an industrial defect segmentation framework based on SAM. Firstly, to bridge the gap between SAM’s pre-training data and distinct characteristics of industrial defects, we introduce a parameter-efficient fine-tuning (PEFT) technique incorporating lightweight Multi-Scale Partial Convolution Aggregation (MSPCA) into Low-Rank Adaptation (LoRA), named MSPCA-LoRA, which effectively enhances the image encoder’s sensitivity to prior knowledge biases, while maintaining PEFT efficiency. Furthermore, we present the Image-to-Prompt Embedding Generator (IPEG), which utilizes image embeddings to autonomously create high-quality prompt embeddings for directing mask segmentation, eliminating the limitations of manually provided prompts. Finally, we apply effective refinements to SAM’s mask decoder, transforming SAM into an end-to-end semantic segmentation framework. On two real-world defect segmentation datasets, PA-SAM achieves mean Intersections over Union of 73.87% and 68.30%, as well as mean Dice coefficients of 84.90% and 80.22%, outperforming other state-of-the-art algorithms, further demonstrating its robust generalization and application potential. Full article

Electricity theft, emerging as one of the severe cyberattacks in smart grids, causes significant economic losses. Due to the powerful expressive ability of deep neural networks (DNN), supervised and unsupervised DNN-based electricity theft detection (ETD) schemes have experienced widespread deployment. However, existing works have the following weak points: Supervised DNN-based schemes require abundant labeled anomalous samples for training, and even worse, cannot detect unseen theft patterns. To avoid the extensively labor-consuming activity of labeling anomalous samples, unsupervised DNNs-based schemes aim to learn the normality of time-series and infer an anomaly score for each data instance, but they fail to capture periodic features effectively. To address these challenges, this paper proposes a novel periodicity-enhanced deep one-class classification framework (PE-DOCC) based on a periodicity-enhanced transformer encoder, named Periodicformer encoder. Specifically, within the encoder, a novel criss-cross periodic attention is proposed to capture both horizontal and vertical periodic features. The Periodicformer encoder is pre-trained by reconstructing partially masked input sequences, and the learned latent representations are then fed into a one-class classification for anomaly detection. Extensive experiments on real-world datasets demonstrate that our proposed PE-DOCC framework outperforms state-of-the-art unsupervised ETD methods. Full article

Deep neural network (DNN)-based classifiers have witnessed great applications in various fields. Unfortunately, the labels of real-world training data are commonly noisy, i.e., the labels of a large percentage of training samples are wrong, which negatively affects the performance of a trained DNN classifier during inference. Therefore, it is challenging to practically formulate a robust DNN classifier using noisy labels in training. To address the above issue, our work designs an effective architecture for training the robust DNN classifier with noisy labels, named a cross dual-branch network guided by meta-data on a single side (CoNet-MS), in which a small amount of clean data, i.e., meta-data, are used to guide the training of the DNN classifier. Specifically, the contributions of our work are threefold. First, based on the principle of small loss, each branch using the base classifier as a neural network module infers partial samples with pseudo-clean labels, which are then used for training another branch through a cross structure that can alleviate the cumulative impact of mis-inference. Second, a meta-guided module is designed and inserted into the single branch, e.g., the upper branch, which dynamically adjusts the ratio between the observed label and the pseudo-label output by the classifier in the loss function for each training sample. The asymmetric dual-branch design makes two classifiers diverge, which facilitates them to filter different types of noisy labels and avoid confirmation bias in self-training. Finally, thorough experiments demonstrate that the trained classifier with our proposal is more robust: the accuracy of the classifier trained with our proposed CoNet-MS on multiple datasets under various ratios of noisy labels and noise types outperforms other classifiers of learning with noisy labels (LNLs), including the state-of-the-art meta-data-based LNL classifier. Full article