Search Results (147)

The work aims to leverage computer vision and artificial intelligence technologies to quantify key components in food catering services. Specifically, it focuses on content identification and portion size estimation in a dining hall setting, typical of corporate and educational settings. An RGB camera is employed to capture the tray delivery process in a self-service restaurant, providing test images for content identification algorithm comparison, using standard evaluation metrics. The approach utilizes the YOLO architecture, a widely recognized deep learning model for object detection and computer vision. The model is trained on labeled image data, and its performance is assessed using a precision–recall curve at a confidence threshold of 0.5, achieving a mean Average Precision (mAP) of 0.873, indicating robust overall performance. The weight estimation procedure combines computer vision techniques to measure food volume using both RGB and depth cameras. Subsequently, density models specific to each food type are applied to estimate the detected food weight. The estimation model’s parameters are calibrated through experiments that generate volume-to-weight conversion tables for different food items. Validation of the system was conducted using rice and chicken, yielding error margins of 5.07% and 3.75%, respectively, demonstrating the feasibility and accuracy of the proposed method. Full article

Recent advancements in food packaging have transitioned from passive containment toward innovative smart systems that integrate active and intelligent functionalities to improve product preservation, safety, and consumer interaction. This review examines the evolution of these technologies, focusing on biodegradable polymers and nanomaterial-enhanced substrates that combine environmental sustainability with superior barriers and antimicrobial performance. Developments in embedded sensing systems, including chemical, temperature, and humidity sensors, enable the continuous monitoring of food quality and environmental conditions, supporting extended shelf-life and early contamination detection. Intelligent packaging further incorporates indicators, sensors, and data carriers that enhance transparency and traceability across supply chains. These systems are often connected through blockchain and Internet of Things (IoT) platforms for real-time data analysis. The review also addresses consumer engagement via interactive labels and personalized nutritional feedback, along with the economic, behavioral, and regulatory aspects influencing large-scale adoption. Life cycle assessments are analyzed to evaluate trade-offs between enhanced functionality and environmental impact, emphasizing recyclability and end-of-life strategies within circular economy frameworks. Finally, the article discusses current technical challenges while highlighting emerging trends such as AI-driven predictive analytics and IoT-enabled connectivity as key enablers of sustainable, efficient, and safe food packaging systems. Full article

This work focuses on finding frequent patterns in continuous flow network traffic Big Data using incremental frequent pattern mining. A newly created Zeek Conn Log MITRE ATT&CK framework labeled dataset, UWF-ZeekData24, generated using the Cyber Range at The University of West Florida, was used for this study. While FP-Growth is effective for static datasets, its standard implementation does not support incremental mining, which poses challenges for applications involving continuously growing data streams, such as network traffic logs. To overcome this limitation, a staged incremental FP-Growth approach is adopted for this work. The novelty of this work is in showing how incremental FP-Growth can be used efficiently on continuous flow network traffic, or streaming network traffic data, where no rebuild is necessary when new transactions are scanned and integrated. Incremental frequent pattern mining also generates feature subsets that are useful for understanding the nature of the individual attack tactics. Hence, a detailed understanding of the features or feature subsets of the seven different MITRE ATT&CK tactics is also presented. For example, the results indicate that core behavioral rules, such as those involving TCP protocols and service associations, emerge early and remain stable throughout later increments. The incremental FP-Growth framework provides a structured lens through which network behaviors can be observed and compared over time, supporting not only classification but also investigative use cases such as anomaly tracking and technique attribution. And finally, the results of this work, the frequent itemsets, will be useful for intrusion detection machine learning/artificial intelligence algorithms. Full article

Early detection of breast cancer commonly relies on imaging technologies such as ultrasound, mammography and MRI. Among these, breast ultrasound is widely used by radiologists to identify and assess lesions. In this study, we developed image segmentation techniques and multiclass classification artificial intelligence (AI) tools based on pretrained models to segment lesions and detect breast cancer. The proposed workflow includes both the development of segmentation models and development of a series of classification models to classify ultrasound images as normal, benign or malignant. The pretrained models were trained and evaluated on the Breast Ultrasound Images (BUSI) dataset, a publicly available collection of grayscale breast ultrasound images with corresponding expert-annotated masks. For segmentation, images and ground-truth masks were used to pretrained encoder (ResNet18, EfficientNet-B0 and MobileNetV2)–decoder (U-Net, U-Net++ and DeepLabV3) models, including the DeepLabV3 architecture integrated with a Frequency-Domain Feature Enhancement Module (FEM). The proposed FEM improves spatial and spectral feature representations using Discrete Fourier Transform (DFT), GroupNorm, dropout regularization and adaptive fusion. For classification, each image was assigned a label (normal, benign or malignant). Optuna, an open-source software framework, was used for hyperparameter optimization and for the testing of various pretrained models to determine the best encoder–decoder segmentation architecture. Five different pretrained models (ResNet18, DenseNet121, InceptionV3, MobielNetV3 and GoogleNet) were optimized for multiclass classification. DeepLabV3 outperformed other segmentation architectures, with consistent performance across training, validation and test images, with Dice Similarity Coefficient (DSC, a metric describing the overlap between predicted and true lesion regions) values of 0.87, 0.80 and 0.83 on training, validation and test sets, respectively. ResNet18:DeepLabV3 achieved an Intersection over Union (IoU) score of 0.78 during training, while ResNet18:U-Net++ achieved the best Dice coefficient (0.83) and IoU (0.71) and area under the curve (AUC, 0.91) scores on the test (unseen) dataset when compared to other models. However, the proposed Resnet18: FrequencyAwareDeepLabV3 (FADeepLabV3) achieved a DSC of 0.85 and an IoU of 0.72 on the test dataset, demonstrating improvements over standard DeepLabV3. Notably, the frequency-domain enhancement substantially improved the AUC from 0.90 to 0.98, indicating enhanced prediction confidence and clinical reliability. For classification, ResNet18 produced an F1 score—a measure combining precision and recall—of 0.95 and an accuracy of 0.90 on the training dataset, while InceptionV3 performed best on the test dataset, with an F1 score of 0.75 and accuracy of 0.83. We demonstrate a comprehensive approach to automate the segmentation and multiclass classification of breast cancer ultrasound images into benign, malignant or normal transfer learning models on an imbalanced ultrasound image dataset. Full article

Deep learning models are increasingly evaluated not only for predictive accuracy but also for their robustness, interpretability, and data quality dependencies. However, current benchmarks largely isolate these dimensions, lacking a unified evaluation protocol that integrates data-centric and model-centric properties. To bridge the gap between data quality assessment and eXplainable Artificial Intelligence (XAI), we introduce DERI1000—the Dataset Explainability Readiness Index—a benchmark that quantifies how suitable and well-prepared a dataset is for explainable and trustworthy deep learning. DERI1000 combines eleven measurable factors—sharpness, noise artifacts, exposure, resolution, duplicates, diversity, separation, imbalance, label noise proxy, XAI overlay, and XAI stability—into a single normalized score calibrated around a reference baseline of 1000. Using five MedMNIST datasets (PathMNIST, ChestMNIST, BloodMNIST, OCTMNIST, OrganCMNIST) and five convolutional neural architectures (DenseNet121, ResNet50, ResNet18, VGG16, EfficientNet-B0), we fitted factor weights through multi-dataset impact analysis. The results indicate that imbalance (0.3319), separation (0.1377), and label noise proxy (0.2161) are the dominant contributors to explainability readiness. Experiments demonstrate that DERI1000 effectively distinguishes models with superficially high accuracy (ACC) but poor interpretability or robustness. The framework thereby enables cross-domain, reproducible evaluation of model performance and data quality under unified metrics. We conclude that DERI1000 provides a scalable, interpretable, and extensible foundation for benchmarking deep learning systems across both data-centric and explainability-driven dimensions. Full article

Printed circuit board (PCB) defect detection is critical to manufacturing quality, yet tiny, low-contrast defects and limited annotations challenge conventional systems. This study develops an ECA–DCN-lite–BiFPN–CARAFE-enhanced YOLOv5 detector by modifying You Only Look Once (YOLO) version 5 (YOLOv5) with Efficient Channel Attention (ECA) for channel re-weighting, a lightweight Deformable Convolution (DCN-lite) for geometric adaptability, a Bi-Directional Feature Pyramid Network (BiFPN) for multi-scale fusion, and Content-Aware ReAssembly of FEatures (CARAFE) for content-aware upsampling. A single-cycle semi-supervised training pipeline is further introduced: a detector trained on labeled images generates high-confidence pseudo-labels for unlabeled data, and the combined set is used for retraining without ratio heuristics. Evaluated on PKU-PCB under label-scarce regimes, the full model improves supervised mean Average Precision at an Intersection-over-Union threshold of 0.5 (mAP@0.5) from 0.870 (baseline) to 0.910, and reaches 0.943 mAP@0.5 with semi-supervision, with consistent class-wise gains and faster convergence. Ablation experiments validate the contribution of each module and identify robust pseudo-label thresholds, while comparisons with recent YOLO variants show favorable accuracy–efficiency trade-offs. These findings indicate that the proposed design delivers accurate, label-efficient PCB inspection suitable for Automated Optical Inspection (AOI) in production environments. This work supports SDG 9 by enhancing intelligent manufacturing systems through reliable, high-precision AI-driven PCB inspection. Full article

Uncrewed aerial systems (UAS)-based remote sensing and artificial intelligence (AI) analysis enable real-time wildfire or bushfire detection, facilitating early response to minimize damage and protect lives and property. However, their effectiveness is limited by three issues: distinguishing smoke from fog, the high cost of manual annotation, and the computational demands of large models. This study addresses the three key challenges by introducing plume as a new indicator to better distinguish smoke from similar visual elements, and by employing a hybrid annotation method using knowledge distillation (KD) to reduce expert labour and accelerate labelling. Additionally, it leverages lightweight YOLO Nano models trained with pseudo-labels generated from a fine-tuned teacher network to lower computational demands while maintaining high detection accuracy for real-time wildfire monitoring. Controlled pile burns in Canungra, QLD, Australia, were conducted to collect UAS-captured images over deciduous vegetation, which were subsequently augmented with the Flame2 dataset, which contains wildfire images of coniferous vegetation. A Grounding DINO model, fine-tuned using few-shot learning, served as the teacher network to generate pseudo-labels for a significant portion of the Flame2 dataset. These pseudo-labels were then used to train student networks consisting of YOLO Nano architectures, specifically versions 5, 8, and 11 (YOLOv5n, YOLOv8n, YOLOv11n). The experimental results show that YOLOv8n and YOLOv5n achieved an mAP@0.5 of 0.721. Plume detection outperforms smoke indicators (F1: 76.1–85.7% vs. 70%) in fog and wildfire scenarios. These findings underscore the value of incorporating plume as a distinct class and utilizing KD, both of which enhance detection accuracy and scalability, ultimately supporting more reliable and timelier wildfire monitoring and response. Full article

To address the challenges of fault diagnosis in wind turbine planetary gearboxes under strong noise and limited labeled target-domain data, this paper proposes a novel intelligent diagnostic method integrating multi-source feature fusion with domain adaptation transfer learning. A Multi-source Feature Attention Fusion Convolutional Neural Network (MSFAF-CNN) is constructed, which dynamically fuses vibration signals from multiple measurement points using a channel attention mechanism to assign optimal weights to the most discriminative features. Furthermore, an improved Multi-source Local Maximum Mean Discrepancy (MS-LMMD) loss is introduced, establishing a hierarchical domain adaptation framework that enables fine-grained alignment of feature distributions between the labeled source and unlabeled target domains. Experimental results under the challenging condition of −4 dB noise demonstrate the superiority of the proposed approach: the cross-condition transfer task (A→B) achieves an accuracy of 95.32%, outperforming the conventional LMMD method by 1.05%. Finally, t-SNE-based visualization confirms that the method enhances cross-domain feature compactness, enabling direct processing of raw vibration signals without manual feature extraction. The findings indicate that the proposed approach offers a highly robust solution for fault diagnosis in drive systems under low signal-to-noise ratios and unlabeled operating conditions. Full article

The automatic identification of P-wave arrival times in microseismic signals is crucial for the intelligent monitoring and early warning of dynamic hazards in coal mines. Traditional methods suffer from low accuracy and poor stability due to complex underground geological conditions and substantial noise interference. This paper proposes a microseismic P-wave arrival time automatic picking model that integrates physical constraints with a deep learning architecture. This study trained and optimized the model using a high-quality, manually labeled dataset. A systematic comparison with the AR picker algorithm and the short-term–long-term average ratio method revealed that this model achieved a precision of 96.60%, a recall of 90.59%, and an F1 score of 93.50% on the test set, with a P-wave arrival time-picking error of less than 20 ms. The average arrival time error was only 5.49 ms, significantly outperforming traditional methods. In cross-mining area generalization tests, the model performed excellently in two mining areas with consistent sampling frequencies (1000 Hz) and high signal-to-noise ratios, demonstrating good engineering transferability. However, its performance decreased in a mining area with a higher sampling rate and stronger noise, indicating its sensitivity to data acquisition parameters. This study developed a high-precision, robust, and potentially cross-domain adaptive model for automatically picking microseismic P-wave arrival times. This model provides support for the automation, precision, and intelligence of coal mine microseismic monitoring systems and has significant practical value in promoting real-time early warning and risk prevention for mine dynamic hazards. Full article

Cyber Threat Intelligence (CTI) reports are essential resources for identifying the Tactics, Techniques, and Procedures (TTPs) of hackers and cyber threat actors. However, these reports are often lengthy and unstructured, which limits their suitability for automatic mapping to the MITRE ATT&CK framework. This study designs and compares five hybrid classification models that combine statistical features (TF-IDF), transformer-based contextual embeddings (BERT and ModernBERT), and topic-level representations (BERTopic) to automatically classify CTI reports into 12 ATT&CK tactic categories. Experiments using the rcATT dataset, consisting of 1490 public threat reports, show that the model integrating TF-IDF and ModernBERT achieved a micro-precision of 72.25%, reflecting a 10.07-percentage-point improvement in detection precision compared with the baseline. The model combining TF-IDF and BERTopic achieved a micro F_0.5 of 67.14% and a macro F_0.5 of 63.20%, demonstrating balanced performance across both frequent and rare tactic classes. These findings indicate that integrating statistical, contextual, and semantic representations can improve the balance between precision and recall while enabling clearer interpretation of model outputs in multi-label CTI classification. Furthermore, the proposed model shows potential applicability for improving detection efficiency and reducing analyst workload in Security Operations Center (SOC) environments. Full article

Anthocyanins, a group of naturally occurring flavonoid compounds, have garnered increasing attention due to their wide-ranging biological activities that suggest their considerable potential to be utilized not only as natural food colorants but also as functional additives that can enhance food preservation and contribute to the development of health-promoting functional foods. Additionally, their sensitivity to environmental factors such as pH and temperature makes anthocyanins promising candidates for use in intelligent packaging systems, particularly as natural indicators for monitoring food freshness and quality throughout storage and distribution. Despite challenges related to their stability and regulatory acceptance, continued research into anthocyanins remains crucial for advancing sustainable, clean-label food technologies and reducing reliance on synthetic additives. To fully leverage their economic and health potential, it is essential to gain a comprehensive understanding of the various plant sources of anthocyanins, their chemical composition, extraction methods, and roles in different applications. Moreover, integrating anthocyanins into food and intelligent packaging systems presents various technical and regulatory challenges that are also summarized in this review. Full article

Medication errors pose a significant threat to patient safety. Although Bar-Code Medication Administration (BCMA) has reduced error rates, it is constrained by handheld devices, workflow interruptions, and incomplete safeguards against wrong patients, wrong doses, or drug incompatibility. In this study, we developed and evaluated a next-generation BCMA system by integrating artificial intelligence and mixed reality technologies for real-time safety checks: Optical Character Recognition verifies medication–label concordance, facial recognition confirms patient identity, and a rules engine evaluates drug–diluent compatibility. Computer vision models achieved high recognition accuracy for drug vials (100%), medication labels (90%), QR codes (90%), and patient faces (90%), with slightly lower performance for intravenous fluids (80%). A mixed-methods evaluation was conducted in a simulated environment using the System Usability Scale (SUS), Reduced Instructional Materials Motivation Survey (RIMMS), Virtual Reality Sickness Questionnaire (VRSQ), and NASA Task Load Index (NASA-TLX). The results indicated excellent usability (median SUS = 82.5/100), strong user motivation (RIMMS = 3.7/5), minimal cybersickness (VRSQ = 0.4/6), and manageable cognitive workload (NASA-TLX = 31.7/100). Qualitative analysis highlighted the system’s potential to streamline workflow and serve as a digital “second verifier.” These findings suggest strong potential for clinical integration, enhancing medication safety at the point of care. Full article

The assessment of wheel–rail force states is a key technical issue in the safety monitoring of high-speed railway turnouts. Due to the complex geometry and severe load fluctuations of turnouts, wheel–rail interactions exhibit strong nonlinearity, asymmetry, and multidimensional coupling characteristics. Traditional methods suffer from limitations such as reliance on labeled samples and poor real-time performance. This study proposes an intelligent evaluation method that integrates multivariate statistical analysis with unsupervised clustering, and establishes a multidimensional analytical framework incorporating data preprocessing, time-domain analysis, safety index evaluation, frequency-domain feature extraction, and cluster-based recognition. Using a turnout section of the Beijing–Tianjin Intercity Railway as a case study, four fundamental wheel–rail force components were selected as feature variables to reveal their dynamic patterns. The DBSCAN density-based clustering algorithm was employed to achieve unsupervised state identification, successfully classifying three typical operating states: normal, high-load abnormal, and extreme load. The clustering silhouette coefficient reached 0.563, significantly outperforming K-means and hierarchical clustering. Safety evaluation results indicate that all relevant indicators meet international standards. The proposed method requires no labeled samples and offers strong physical interpretability and engineering applicability, providing effective support for turnout condition awareness and predictive maintenance. Full article

This research contrasts model-centric (MCAI) and data-centric (DCAI) strategies in artificial intelligence, focusing specifically on optical quality control. It addresses the necessity for a thorough empirical study to evaluate both approaches under identical conditions. By examining casting and leather datasets, the study highlights that the quality and diversity of data play a more vital role in the success of models than merely fine-tuning hyperparameters. While MCAI delivers dependable results with superior datasets, DCAI methods—such as label correction, data augmentation, and generating synthetic data through diffusion models—significantly enhance recognition performance. For the casting dataset, accuracy increased from 83% to 93%, and for the leather dataset, from 53% to 62%. These results indicate that robust AI systems are built on high-quality, balanced data. Full article

To address the problems of noise sensitivity, insufficient modeling of long-term time-series dependence, and high cost of labeled data in the fault diagnosis of pumping station units, an intelligent diagnosis method integrating the improved Sparrow Search Algorithm (tSSA) and Informer model is proposed in this study. Firstly, an adaptive t-distribution strategy is introduced into the Sparrow Search Algorithm to dynamically adjust the degree of freedom parameters of the mutation operator, balance global search and local development capabilities, avoid the algorithm converging to the origin, and enhance optimization accuracy, with time complexity consistent with the original SSA. Secondly, by combining the sparse self-attention and self-attention distillation mechanisms of Informer, the model’s ability to extract key features of long sequences is optimized, and its hyperparameters are adaptively optimized via tSSA. Experiments were conducted based on 12 types of fault vibration data acquired from pumping station units. Outliers were removed using the interquartile range (IQR) method, and dimensionality reduction was achieved through kernel principal component analysis (KPCA). The results indicate that the average diagnostic accuracy of tSSA-Informer under noise-free conditions reaches 98.73%, which is significantly higher than that of models such as SSA-Informer and GA-Informer; under noise interference of SNR = −1 dB, it still maintains an accuracy of 87.47%, outperforming comparative methods like 1D-DCTN; when the labeled sample size is reduced to 10%, its accuracy is 61.32%, which is more than 40% higher than that of traditional models. These results verify the robustness and practicality of the proposed method in strong-noise and small-sample scenarios. This study provides an efficient solution for the intelligent fault diagnosis of complex industrial equipment. Full article