Search Results (263)

Open-world object detection (OWOD) aims to build detectors that can recognize known categories while simultaneously identifying unknown objects and incrementally learning novel classes. Despite recent advances, existing OWOD approaches still struggle with two critical challenges: the severe bias toward head classes in long-tailed data distributions and the misclassification of unknown objects as background. To address these issues, we introduce TAPM (Tail-Calibrated Transformer Autoencoding with Prototype-Guided Mining), a novel framework that explicitly enhances tail-class representation and robustly reveals unknown objects. TAPM integrates three core innovations: (1) a transformer-based autoencoder that reconstructs region features to calibrate embeddings for rare categories, mitigating the dominance of frequent classes; (2) a prototype-guided mining strategy that leverages class prototypes to localize both overlooked tail instances and candidate unknowns; and (3) an uncertainty-aware soft-labeling mechanism that assigns probabilistic supervision to pseudo-unknowns, reducing noise in incremental learning. Extensive experiments on the MS-COCO and LVIS benchmarks demonstrate that TAPM significantly improves unknown-object recall while maintaining strong known-class accuracy, achieving state-of-the-art performance across both the superclass-separated (S-OWODB) and superclass-mixed (M-OWODB) benchmarks. In particular, TAPM achieves a +20.4-point gain in U-Recall over the strong PROB baseline, underscoring its effectiveness in detecting novel objects without sacrificing mean Average Precision (mAP). Furthermore, TAPM achieves better generalization on cross-dataset evaluations, highlighting its robustness in diverse open-world scenarios. Full article

In this paper, the impact of SiN passivation on dynamic-R_ON degradation of AlGaN/GaN HEMTs devices is put in evidence. To this end, samples showing different SiN passivation stoichiometry are considered, labeled as Sample A and Sample B. For dynamic-R_ON tests, two different experimental setups are employed to investigate the R_ON-drift showing up during conventional switch mode operation by driving the DUTs under both (i) resistive load and (ii) soft-switching trajectory. This allows to discern the impact of hot carriers and off-state drain voltage stress on the R_ON parameter drift. Measurements performed with both switching loci shows similar dynamic-R_ON response, indicating that hot carriers are not involved in the degradation of tested devices. Nevertheless, a significant difference was observed between Sample A and Sample B, with the former showing an additional R_ON-degradation mechanism, not present on the latter. This additional drift is totally ascribed to the SiN passivation layer and is confirmed by the different leakage current measured across the two SiN types. The mechanism is explained by the injection of negative charges from the Source Field-Plate towards the AlGaN surface that are captured by surface/dielectric states and partially depletes the 2DEG underneath. Full article

Granular jamming grippers have emerged as a versatile solution in soft robotics due to their ability to manipulate objects of various shapes and sizes, earning them the label of “universal grippers”. They are composed of granular material confined within an elastic membrane that conforms to the object like a fluid and solidifies upon vacuum application, enabling a firm grip through friction and grain interlocking. This work provides a systematic review of the state of the art, addressing their physical principles, the influence of grain and membrane properties, performance characterization methods, and applications across diverse fields. Additionally, the main control variables of these grippers closely related to state variables used in control systems are discussed, along with the current knowledge gaps. Finally, five potential directions for future research are proposed. Full article

Industrial process monitoring increasingly relies on sensor-generated time-series data, yet the lack of labels, high variability, and operational noise make it difficult to extract meaningful patterns using conventional methods. Existing clustering techniques either rely on fixed distance metrics or deep models designed for static data, limiting their ability to handle dynamic, unstructured industrial sequences. Addressing this gap, this paper proposes a novel framework for unsupervised discovery of operational modes in univariate time-series data using image-based convolutional clustering with composite internal evaluation. The proposed framework improves upon existing approaches in three ways: (1) raw time-series sequences are transformed into grayscale matrix representations via overlapping sliding windows, allowing effective feature extraction using a deep convolutional autoencoder; (2) the framework integrates both soft and hard clustering outputs and refines the selection through a two-stage strategy; and (3) clustering performance is objectively evaluated by a newly developed composite score, $S_{eva}$, which combines normalized Silhouette, Calinski–Harabasz, and Davies–Bouldin indices. Applied to over 3900 furnace melting operations from a Nordic foundry, the method identifies seven explainable operational patterns, revealing significant differences in energy consumption, thermal dynamics, and production duration. Compared to classical and deep clustering baselines, the proposed approach achieves superior overall performance, greater robustness, and domain-aligned explainability. The framework addresses key challenges in unsupervised time-series analysis, such as sequence irregularity, overlapping modes, and metric inconsistency, and provides a generalizable solution for data-driven diagnostics and energy optimization in industrial systems. Full article

This study aimed to (i) investigate the effect of using grape water in the production of traditional sourdough; (ii) select seeds for use in laboratory-scale sourdough bread production; and (iii) assess the effect of incorporating fresh germinated seeds into recipe of organic sourdough bread on nutritional, technological, and sensory properties. The pH of both control (CSD, flour only) and boosted (BSD, supplemented with “grape water”) sourdough fell below 4.5 by day 3. After 10 days of back-slopping and fermentation, both sourdoughs harbored 9 log CFU/g of lactic acid bacteria, whereas yeast cell density in the CSD was 1 log cycle higher. Based on their high germination rates (~90%), lentil and wheat seeds were selected as additional ingredients (5%). Bread with germinated lentils (GL) and bread with germinated wheat (GW) were compared with control bread (without seeds). GL and GW breads showed gas cell areas of 28.6% and 18.1%, respectively, which were higher than the control. In addition, GL and GW received higher scores for taste (8.6) and softness (5.6), respectively. Additionally, GL contained more proteins (9.9%) and fewer lipids (0.3%) than the two other bread types, in addition to being potentially labeled as a “source of fiber”. Full article

Background/Objectives: The nasal bone is critical to both the functional integrity and esthetic contour of the facial skeleton. Nasal bone fractures constitute the most prevalent facial fracture presentation in emergency departments. The identification of these fractures and the determination of immediate intervention requirements pose significant challenges for inexperienced residents, potentially leading to oversight. Methods: A retrospective analysis was conducted on facial trauma patients undergoing cranial radiography (Waters’ view) during initial emergency department assessment between March 2008 and July 2022. This study incorporated 2099 radiographic images. Surgical indications comprised the displacement angle, interosseous gap size, soft tissue swelling thickness, and subcutaneous emphysema. A deep learning-based artificial intelligence (AI) algorithm was designed, trained, and validated for fracture detection on radiographic images. Model performance was quantified through accuracy, precision, recall, and F1 score. Hyperparameters included the batch size (20), epochs (70), 50-layer network architecture, Adam optimizer, and initial learning rate (0.001). Results: The deep learning AI model employing segmentation labeling demonstrated 97.68% accuracy, 82.2% precision, 88.9% recall, and an 85.4% F1 score in nasal bone fracture identification. These outcomes informed the development of a predictive algorithm for guiding conservative versus surgical management decisions. Conclusions: The proposed AI-driven algorithm and criteria exhibit high diagnostic accuracy and operational efficiency in both detecting nasal bone fractures and predicting surgical indications, establishing its utility as a clinical decision-support tool in emergency settings. Full article

This review article provides a comprehensive evaluation of Infuse^® and InductOs^®, two ground-breaking recombinant human Bone Morphogenetic Protein-2 (rhBMP-2)-based bone graft products, focusing on their tissue-level regenerative responses, clinical applications, and associated costs. Preclinical and clinical studies demonstrate that rhBMP-2 induces strong osteoinductive activity, effectively promoting mesenchymal stem cell differentiation and vascularized bone remodeling. While generally well-tolerated, these osteoinductive effects are dose-dependent, and excessive dosing or off-label use may result in adverse outcomes, such as ectopic bone formation or soft tissue inflammation. Histological and imaging analyses in craniofacial, orthopedic, and spinal fusion models confirm significant bone regeneration, positioning rhBMP-2 as a viable alternative to autologous grafts. Notably, advances in delivery systems and scaffold design have enhanced the stability, bioavailability, and targeted release of rhBMP-2, leading to improved fusion rates and reduced healing times in selected patient populations. These innovations, alongside its proven regenerative efficacy, underscore its potential to expand treatment options in cases where autografts are limited or unsuitable. However, the high initial cost, primarily driven by rhBMP-2, remains a critical limitation. Although some studies suggest overall treatment costs might be comparable to autografts when factoring in reduced complications and operative time, autografts often remain more cost-effective. Infuse^® has not substantially reduced the cost of bone regeneration and presents additional safety concerns due to the rapid (burst) release of growth factors and limited mechanical scaffold support. Despite representing a significant advancement in synthetic bone grafting, further innovation is essential to overcome limitations related to cost, mechanical properties, and controlled growth factor delivery. Full article

A physics-informed neural network (PINN) framework is developed to model the large deformation and coupled electromechanical response of dielectric elastomer tubes for energy harvesting. The system integrates incompressible neo-Hookean elasticity with radial electric loading and compressible gas inflation, leading to nonlinear equilibrium equations with deformation-dependent boundary conditions. By embedding the governing equations and boundary conditions directly into its loss function, the PINN enables accurate, mesh-free solutions without requiring labeled data. It captures realistic pressure–volume interactions that are difficult to address analytically or through conventional numerical methods. The results show that internal volume increases by over 290% during inflation at higher reference pressures, with residual stretch after deflation reaching 9.6 times the undeformed volume. The axial force, initially tensile, becomes compressive at high voltages and pressures due to electromechanical loading and geometric constraints. Harvested energy increases strongly with pressure, while voltage contributes meaningfully only beyond a critical threshold. To ensure stable training across coupled stages, the network is optimized using the Optuna algorithm. Overall, the proposed framework offers a robust and flexible tool for predictive modeling and design of soft energy harvesters. Full article

The widespread adoption of smartphones and smartwatches has enabled non-intrusive fall detection through built-in sensors and on-device computation. While these devices are widely used by older adults, existing systems still struggle to accurately detect soft falls in real-world settings. There is a notable drop in performance when fall-detection models trained offline on labeled accelerometer data are deployed and tested in real-world conditions using streaming, real-time data. To address this, our experimental study investigates whether incorporating additional sensor modalities, specifically gyroscope data with accelerometer data from wrist and hip locations, can help bridge this performance gap. Through systematic experimentation, we demonstrated that combining accelerometer data from the hip and the wrist yields a model capable of achieving an F1-score of 88% using a Transformer-based neural network in offline evaluation, which is an improvement of 8% over a model trained solely on wrist accelerometer data. However, when it is deployed in an uncontrolled home environment with streaming real-time data, this model produced a high number of false positives. To address this, we retrained the model using feedback data that comprised both false positives and true positives and was collected from ten participants during real-time testing. This refinement yielded an F1-sore of 92% and significantly reduced false positives while maintaining comparable accuracy in detecting true falls in real-world settings. Furthermore, we demonstrated that the improved model generalizes well to older adults’ movement patterns, with minimal false-positive detections. Full article

Meta-learning has demonstrated significant advantages in small-sample tasks and has attracted considerable attention in wind turbine fault diagnosis. However, due to extreme operating conditions and equipment aging, the monitoring data of wind turbines often contain false alarms or missed detections. This results in inaccurate fault sample labeling. In meta-learning, these erroneous labels not only fail to help models quickly adapt to new meta-test tasks, but they also interfere with learning for new tasks, which leads to “negative transfer” phenomena. To address this, this paper proposes a novel method called Online Soft-Labeled Meta-learning with Gaussian Prototype Networks (SL-GPN). During training, the method dynamically aggregates feature similarities across multiple tasks or samples to form online soft labels. They guide model training process and effectively solve small-sample bearing fault diagnosis challenges. Experimental tests on small-sample data under various operating conditions and error labels were carried out. The results show that the proposed method improves diagnostic accuracy in small-sample environments, reduces false alarm rates, and demonstrates excellent generalization performance. Full article

In recent years, the proliferation of fake news and misinformation has grown exponentially, far surpassing that of genuine news and posing a serious threat to social stability. Existing research in fake news detection primarily applies contrastive learning methods with a single-hot labeling strategy. The issue does not lie with contrastive learning as a technique but with its current application in fake news detection systems. Specifically, these systems penalize all negative samples equally due to the use of single-hot labeling, thus overlooking the underlying semantic relationships among negative samples. As a result, contrastive learning models tend to learn from simple samples while neglecting highly deceptive samples located at the boundary between true and false, as well as the heterogeneity of text-image features, which complicates cross-modal fusion. To mitigate these known limitations in current applications, this paper proposes a fake news detection method based on contrastive learning and cross-modal interaction. First, a consistency-aware soft-label contrastive learning mechanism based on semantic similarity is designed to provide more granular supervision signals for contrastive learning. Secondly, a difficult negative sample mining strategy based on a similarity matrix is designed to optimize the symmetry alignment of image and text features, which effectively improves the model’s ability to discriminate boundary samples. To further optimize the feature fusion process, a cross-modal interaction module is designed to learn the symmetric interaction relationship between image and text features. Finally, an attention mechanism is designed to adaptively adjust the contributions of text-image features and interaction features, forming the final multimodal feature representation. Experiments are conducted on two major social media platform datasets, and compared with existing methods, the proposed method effectively improves the detection capability of fake news. Full article

Most public seismic recordings, sampled at hundreds of Hz, tend to be unlabelled, i.e., not catalogued, mainly because of the sheer volume of samples and the amount of time needed by experts to confidently label detected events. This is especially challenging for very low signal-to-noise ratio microseismic events that characterise landslides during rock and soil mass displacement. Whilst numerous supervised machine learning models have been proposed to classify landslide events, they rely on a large amount of labelled datasets. Therefore, there is an urgent need to develop tools to effectively automate the data-labelling process from a small set of labelled samples. In this paper, we propose a semi-supervised method for labelling of signals recorded by seismometers that can reduce the time and expertise needed to create fully annotated datasets. The proposed Siamese network approach learns best class-exemplar anchors, leveraging learned similarity between these anchor embeddings and unlabelled signals. Classification is performed via soft-labelling and thresholding instead of hard class boundaries. Furthermore, network output explainability is used to explain misclassifications and we demonstrate the effect of anchors on performance, via ablation studies. The proposed approach classifies four landslide classes, namely earthquakes, micro-quakes, rockfall and anthropogenic noise, demonstrating good agreement with manually detected events while requiring few training data to be effective, hence reducing the time needed for labelling and updating models. Full article

Unsupervised feature selection aims to characterize the cluster structure of original features and select the optimal subset without label guidance. However, existing methods overlook fuzzy information in the data, failing to model cluster structures between data effectively, and rely on squared error for data reconstruction, exacerbating noise impact. Therefore, a robust unsupervised feature selection algorithm based on fuzzy anchor graphs (FWFGFS) is proposed. To address the inaccuracies in neighbor assignments, a fuzzy anchor graph learning mechanism is designed. This mechanism models the association between nodes and clusters using fuzzy membership distributions, effectively capturing potential fuzzy neighborhood relationships between nodes and avoiding rigid assignments to specific clusters. This soft cluster assignment mechanism improves clustering accuracy and the robustness of the graph structure while maintaining low computational costs. Additionally, to mitigate the interference of noise in the feature selection process, an adaptive fuzzy weighting mechanism is presented. This mechanism assigns different weights to features based on their contribution to the error, thereby reducing errors caused by redundant features and noise. Orthogonal tri-factorization is applied to the low-dimensional representation matrix. This guarantees that each center represents only one class of features, resulting in more independent cluster centers. Experimental results on 12 public datasets show that FWFGFS improves the average clustering accuracy by 5.68% to 13.79% compared with the state-of-the-art methods. Full article

Class-incremental learning (CIL) in deep neural networks is affected by catastrophic forgetting (CF), where acquiring knowledge of new classes leads to the significant degradation of previously learned representations. This challenge is particularly severe in medical image analysis, where costly, expertise-dependent annotations frequently contain pervasive and hard-to-detect noisy labels that substantially compromise model performance. While existing approaches have predominantly addressed CF and noisy labels as separate problems, their combined effects remain largely unexplored. To address this critical gap, this paper presents a dual-stage clean-sample selection method for Incremental Noisy Label Learning (DSCNL). Our approach comprises two key components: (1) a dual-stage clean-sample selection module that identifies and leverages high-confidence samples to guide the learning of reliable representations while mitigating noise propagation during training, and (2) an experience soft-replay strategy for memory rehearsal to improve the model’s robustness and generalization in the presence of historical noisy labels. This integrated framework effectively suppresses the adverse influence of noisy labels while simultaneously alleviating catastrophic forgetting. Extensive evaluations on public medical image datasets demonstrate that DSCNL consistently outperforms state-of-the-art CIL methods across diverse classification tasks. The proposed method boosts the average accuracy by 55% and 31% compared with baseline methods on datasets with different noise levels, and achieves an average noise reduction rate of 73% under original noise conditions, highlighting its effectiveness and applicability in real-world medical imaging scenarios. Full article

This study proposes a deep metric learning-based pavement distress classification method to address critical limitations in conventional approaches, including their dependency on large training datasets and inability to incrementally learn new categories. To resolve high intra-class variance and low inter-class distinction in distress images, we design a CNN head with multi-cluster centroins trained via SoftTriple loss, simultaneously maximizing inter-class separation while establishing multiple intra-class centers. An adaptive weighting strategy combining sample similarity and class priors mitigates data imbalance, while soft-label techniques reduce labeling noise by evaluating similarity against support-set exemplars. Evaluations on the UAV-PDD2023 dataset demonstrate superior performance—3.2% higher macro-recall than supervised learning, and 6.7%/8.5% improvements in macro-F1/weighted-F1 over iCaRL incremental learning—validating the method’s effectiveness for real-world road inspection scenarios with evolving distress types and limited annotation. Full article