Search Results (686)

Wine labels influence consumer purchasing choices. With no- and lower-alcohol wines increasingly available, and marketing often centering on healthier choices, it is important to consider the information provided on the labels of these products. Therefore, this cross-sectional audit of no- and lower-alcohol wines presents a snapshot representation of earlier emerging data found in three major retail outlets on the Central Coast, NSW, Australia, that were audited for front and back label information (including alcohol content, nutrition, and health messaging). One hundred and eleven unique labels were identified (46 no-alcohol and 65 lower-alcohol wines). Most lower-alcohol wines (72.3%) typically included information on alcohol by volume ABV% on front labels. Narrative wording of alcohol content on front labels was more frequent (81.5%); however, wording with “Lighter in alcohol” was 50%, followed by “Lower alcohol” at 12.3%. No-alcohol wines reported ABV% on less than half of front labels (43.5%), but all front labels included a narrative descriptor of alcohol content. The most common descriptor was “Zero alcohol” (41.3%), followed by “Alcohol free” (23.9%). All no-alcohol wines and 44.6% of the lower-alcohol wines contained nutrition information panels. Almost half the back labels included a diversity of marketing messaging, with overall themes of suggested drinking occasion and/or health-related properties (43.1% lower-alcohol labels: 45.7% no-alcohol labels). These data show a variance in alcohol content labeling, associated descriptors, and presentation of nutritional and other information. This study is important for directing future research on the utility and understanding of these labels. Full article

Rock fissure parameters, such as length, width, and density, are essential for analyzing the progressive instability of open-pit mine slopes. Under the combined effects of engineering disturbance, geological conditions, and environmental factors, slope fissures continuously propagate and evolve. However, large variations in fissure scale, complex rock-surface textures, blurred boundaries, and weak micro-fissure features increase the difficulty of intelligent fissure segmentation, identification, and parameter extraction. Consequently, many mining enterprises still rely on manual interpretation, which is time-consuming and susceptible to subjective errors. To address these challenges, this study develops Crack-YOLO, a task-oriented fissure detection and instance-segmentation model based on YOLOv8-Seg. A total of 500 original UAV images were collected from multiple open-pit mines and processed to construct a dataset containing 3600 fissure image patches, including 3240 images for training and 360 images for testing. In Crack-YOLO, selected C2f modules are replaced with contextual semantic enhancement modules (CoT Blocks), and a texture information enhancement module (SM Block) is incorporated to strengthen contextual semantic representation and fine-grained texture-feature extraction. The model achieved segmentation precision, recall, mAP50, and mAP50:95 values of 0.896, 0.787, 0.854, and 0.392, respectively. For object detection, the corresponding values were 0.968, 0.862, 0.959, and 0.773, respectively. The segmentation results were further processed using K3M skeleton extraction and physical-scale calibration to quantitatively extract geometric parameters, including fissure length, equivalent average width, and azimuth. Validation using an image containing seven representative fissures yielded mean absolute errors of 0.016 m, 0.010 m, and 0.90° for fissure length, equivalent average width, and azimuth, respectively, indicating the feasibility of the proposed parameter-quantification workflow. In an application test conducted in a typical open-pit mine scene, the proposed workflow identified 196 fissures within approximately 22 s and quantitatively analyzed their geometric parameters and distribution characteristics. The results indicate that the proposed method has potential for fissure identification and geometric-parameter quantification in open-pit mine slopes and may provide quantitative data support for slope-fissure monitoring and stability analysis. Full article

Coral reef degradation has accelerated the development of active restoration approaches, including the use of mid-water coral nurseries (Floating Reef Devices; FRDs) designed to enhance coral growth and survival. Although primarily used for coral propagation, the broader ecological roles of FRDs remain poorly understood. We assess the biodiversity and ecological structure of a long-term FRD deployed in the Gulf of Eilat over two decades as an active coral nursery. We surveyed naturally recruited sessile organisms and associated fish communities, quantifying taxonomic breadth and distinctness (Δ+), functional diversity, and trophic structure. The FRD supported a diverse community spanning eight phyla and 128 genera, with high taxonomic distinctness and wide functional trait representation across multiple trophic levels. Despite lower genus richness compared to natural habitats, coral and fish assemblages exhibited taxonomic and functional patterns that fell within the range observed for the adjacent natural reef. Functional analyses showed that the FRD’s coral assemblage occupied nearly the full regional trait space, while fish biomass was dominated by planktivores and secondary consumers, indicating a differently structured trophic organization. These findings demonstrate that long-term floating nursery systems can develop into structurally complex, multi-trophic communities, providing ecological and conservation values and ecosystem services beyond their original restoration role. These findings provide an ecological foundation for future studies of floating restoration systems and their associated ecological roles, while also informing their design, implementation, and evaluation as practical tools for enhancing ecosystem recovery and associated ecological functions. Full article

The number of grains per spike is a critical determinant of wheat yield and plays an important role in phenotyping, breeding evaluation, and yield-related trait analysis. However, conventional grain counting methods rely on manual operation, which is time-consuming, labor-intensive, and prone to subjective bias, making them unsuitable for high-throughput applications. To address these limitations, this study proposes an improved wheat grain detection and counting method based on YOLOv8 integrated with a P2 detection layer and an Efficient Multi-Scale Attention (EMA) mechanism. The P2 detection layer enhances the resolution of shallow feature maps, improving the model’s ability to detect small and densely distributed grains. Meanwhile, the EMA module strengthens multi-scale feature representation and improves training stability and generalization performance, particularly in complex canopy and overlapping grain scenarios. Experimental results demonstrate that the proposed YOLOv8-EMA-P2 model achieves a precision of 96.10%, recall of 95.60%, mAP@0.5 of 96.80%, and mAP@0.5:0.95 of 68.40% on the test set, indicating strong detection performance. For counting performance, the model achieves a coefficient of determination (R²) of 0.8384, with a root mean square error (RMSE) of 1.8517, mean absolute error (MAE) of 1.2628, and mean relative error (MRE) of 5.33%. In addition, the average inference time per image is 6.5 ms, demonstrating its potential for real-time applications. Overall, the proposed method improves the accuracy, robustness, and efficiency of wheat grain detection and counting, providing an effective solution for automated wheat phenotyping and yield estimation. Full article

Electric motors consume over 45% of global electricity and are a primary source of unplanned industrial downtime. Real-time fault detection at scale faces severe constraints, including distributed topologies, intermittent connectivity, and strict energy budgets on battery-powered edge nodes. Existing hierarchical federated learning approaches address resource disparities across tiers but lack downward feature transfer. This prevents resource-constrained edge sensors from utilizing cloud-learned representations when local fault data is sparse. This paper proposes a hierarchical federated cyber-physical architecture featuring three online cross-layer transfer mechanisms: warm-start Convolutional Neural Network (CNN) weight extraction, Long Short-Term Memory (LSTM) embedding alignment, and adaptive teacher–student distillation. This work is best characterized as a hierarchical federated-learning protocol and edge-hardware feasibility study: it validates the communication protocol, cross-layer transfer mechanisms, and sensor-tier hardware budget end-to-end, using the Gym-Electric-Motor (GEM) simulator as a controlled, reproducible, and openly available substitute for physically instrumented motor faults, rather than as a validated physical motor-fault diagnosis system. The framework is evaluated on a ten-node low-power System-on-Chip (SoC) microcontroller, low-power single-board computer, and cloud computing platform test bench, using GEM-simulated operating trajectories as a controlled, reproducible proxy for non-IID motor fault conditions in five-class motor fault detection. Under this test bench, the framework achieves an over 3-fold convergence speedup and reduces the sensor–cloud accuracy gap by nearly 74% (from 10.1% down to 2.7%) at a 200× lower compute budget. It improves minority-class diagnostic reliability, with F1 scores improving by 45% on average, while achieving over 95% sensor accuracy on this test bench, and the proposed mechanism also limits Macro-F1 degradation under injected sensor noise (SNR = 10 dB) to 8.0%, versus up to 22.5% for independent per-tier training. From an embedded electronics implementation perspective, the system operates under a 3.3 ms latency and consumes only 2.1 mJ per inference on the low-power SoC hardware—outperforming prior edge PdM deployments that report 4–6 mJ per inference—indicating a feasible architecture for energy-constrained edge intelligence, pending validation on physically measured fault data. Full article

The expansion of the gluten-free food market is driven by two distinct factors: medical prescriptions for people with celiac disease, and lifestyle adoption among non-celiac consumers. This study examines how the use of the English term “gluten-free” may influence the cognitive conceptualization of the product compared to the term “libre de gluten” among two groups. Methodologically, the cognitive structure of 240 Mexican consumers (120 with celiac disease and 120 without) was evaluated using social representation theory and the word association technique. Findings reveal that the language used on labels significantly influences the semantic categories evoked. The non-celiac group exhibits an intellectualized and stable representation for both variants, prioritizing health and composition. In contrast, the celiac group experiences a dichotomy: the Spanish version triggers a restrictive reality linked to the consumption experience and economic factors, while Anglicism may shift the perception towards health and aspirational dimensions. It is concluded that the linguistic framework of the labeling operates as a distinct cultural vehicle; Anglicism denotes status for the non-celiac consumer but may evoke an aspirational mechanism that mitigates disease barriers in the celiac patient. Full article

Reliable bearing fault detection is essential for predictive maintenance in industrial systems; however, obtaining labelled fault data is often expensive, time-consuming, and impractical in real-world deployments. To address this challenge, this study proposes a healthy-only self-supervised anomaly detection framework for bearing health monitoring using vibration measurements. The proposed approach combines convolutional neural networks and Transformer-based temporal modelling to learn informative representations from healthy vibration signals without requiring fault labels during representation learning. Three self-supervised learning strategies—reconstruction-based, contrastive, and a unified contrastive–reconstruction objective—are investigated to evaluate the effectiveness of different representation learning approaches. The learned latent representations are subsequently analysed using Isolation Forest and Mahalanobis-distance anomaly scoring methods. To provide a realistic assessment of generalisation, a strict grouped cross-validation protocol is employed, where data are partitioned at the sample level to prevent information leakage between training and testing sets. Furthermore, prevalence-aware experiments are conducted under 5% and 10% fault prevalence scenarios to assess deployment robustness. Experimental results on the Paderborn bearing dataset demonstrate that the proposed CNN + Transformer model trained with combined contrastive and reconstruction objectives and evaluated using Isolation Forest achieves the best overall performance, obtaining a ROC-AUC of $0.878\pm 0.015$, a PR-AUC of $0.958\pm 0.005$, and an F1-score of $0.590\pm 0.040$. The results consistently outperform classical feature-based approaches, One-Class SVM, and autoencoder baselines. Ablation analysis further shows that combining contrastive and reconstruction objectives produces more informative representations than either objective alone. The findings demonstrate that the proposed healthy-only self-supervised framework provides an effective and label-efficient approach for rolling bearing anomaly detection and shows promise for predictive maintenance applications where labelled fault data are limited or unavailable. Full article

This paper presents a hybrid computer vision framework that explicitly decouples geometric stem measurement from visual foliage condition classification for automated post-harvest quality assessment of Ruscus hypophyllum ornamental foliage. The proposed approach addresses a gap in the literature where heterogeneous quality attributes are typically treated within a single unified learning framework. In the first stage, stem size is estimated using a pixel-based geometric method that incorporates trigonometric orientation correction and spatial calibration via a reference marker of known length, enabling accurate conversion of image measurements to real-world physical dimensions. In the second stage, foliage condition is classified as good or poor using supervised machine learning models trained on Bag of Features representations extracted with the SIFT descriptor. A dataset of 1233 Ruscus hypophyllum images was acquired under controlled conditions using a consumer-grade smartphone camera and processed using open-source Python 3.11 libraries. Twenty-four classifier configurations across six model families were evaluated using stratified 10-fold cross-validation. The geometric estimation stage achieved a Mean Absolute Error (MAE) of 1.2 mm, a Root Mean Square Error (RMSE) of approximately 1.3 mm, and a size categorization accuracy of 99.84%. For foliage condition classification, the Linear Support Vector Machine achieved the best performance, with an accuracy of $92.4\pm 1.0$% and an F1-score of $91.1\pm 1.2$%, outperforming all other evaluated configurations. The proposed framework provides an interpretable, computationally efficient, and accessible solution for automated foliage quality grading, with potential applications in export-oriented ornamental foliage processing facilities. Full article

Posture transition behavior in sows approaching parturition provides an important physiological cue for farrowing prediction. However, manual monitoring is time-consuming, labor-intensive and difficult to sustain under nighttime production conditions, while existing machine vision approaches remain limited in their ability to represent continuous posture transitions in complex farm environments. Here, we propose an event-level posture transition recognition framework that integrates YOLOv11n with an RGB–Flow cross-modal temporal network. YOLOv11n is first used to detect basic sow postures at the frame level, after which candidate transition events are automatically generated and refined according to temporal state changes. For each event segment, RGB appearance features and optical-flow motion features are extracted to construct dual-branch spatio-temporal representations. We further develop a multi-scale cross-modal attention temporal network (MS-CMATNet) for event-level behavior classification. The network captures local temporal dynamics through a multi-scale module, enhances interactions between RGB and Flow representations through cross-modal attention, and improves feature discriminability and stability by incorporating temporal–channel attention blocks (TCBAM) and an auxiliary cross-modal consistency loss (AuxCross). Experiments show that MS-CMATNet achieves an Accuracy of 88.14%, a Macro-Recall of 84.04%, and a Weighted-F1 score of 87.66% under the fixed training/validation split, outperforming the compared machine learning models, deep temporal models, and representative temporal and cross-modal baselines. Repeated stratified cross-validation and paired t-tests further confirm that MS-CMATNet achieves statistically reliable improvements over most compared baselines, particularly in Macro-F1 and Weighted-F1. These findings demonstrate the potential of the proposed framework for automated farrowing prediction in smart livestock farming. Full article

Movement-based and physical activity programs are central tools in autism intervention, so recognizing the activities a child performs during therapy is valuable for objective progress tracking. Manual monitoring of these sessions is time-consuming and subjective, and raw videos raise privacy concerns because it shows identifiable children. We address autism therapeutic activity recognition from privacy-preserving 2D skeletons, and we focus on the practical difficulty of how several therapeutic activities differ only in subtle motion details. As a backbone, we adopt ProtoGCN, a graph convolutional network that represents each action as a combination of learnable motion prototypes. However, this contrastive backbone organizes all classes at once, so it does not enforce a margin between the few pairs that remain entangled after training. We therefore introduce a Refine–Confusable (RC) module, a training-only regularizer that pushes apart the empirically most-confused class pairs using a hinge-margin loss over momentum-updated class centroids. The module changes neither the backbone nor the inference cost. On the MMASD dataset, restricted to the ten-class 2D-skeleton configuration, the RC module improves the base model across random, session-independent, and subject-independent evaluation. The gain is largest on the strictest subject-independent split and a clip-level analysis confirms that this improvement is statistically significant. Under the protocol-matched holdout, the method reaches 96.30% accuracy with 0.959 macro-F1, surpassing recent 2D-skeleton baselines while keeping a lightweight and privacy-preserving modality. The improvements are modest, as expected on a small clinical dataset, and t-SNE and prototype visualizations show that the learned representation is discriminative and interpretable. Full article

The Amish are among the most recognizable religious minorities in the United States. Their religious convictions lead them to refuse participation in filmmaking, yet they have become a subject of sustained interest in cinema. To date, more than forty films have featured the Amish as their primary subject or a significant narrative backdrop, with three-quarters of these appearing in the twenty-first century. Drawing on close reading and content analysis of this complete corpus, spanning from 1955 to 2025, Amish society is repeatedly rendered on screen as an “Arcadian” ideal: vast and tranquil pastoral landscapes, a simple and unhurried life of self-sufficiency, and warm, close-knit families and communities. While individual films vary in genre, style, and narrative focus, and the overall tone has grown lighter and more consumer-oriented in the twenty-first century, the Arcadian character of this image has remained largely intact across seven decades of filmmaking. The Amish image on screen is, however, largely romanticized, diverging considerably from the complex realities of Amish life. Because the Amish do not participate in filmmaking or public self-representation, their image has been shaped almost entirely from the outside—by a wider society that has long projected its anxieties about modern life and its longings for a simpler world onto this community, and by a film industry that has gradually settled into a fixed set of narrative conventions. Full article

Smartphones incorporate acoustic components, including a speaker and multiple microphones, which can be used as a low-cost, contactless platform for vital signs monitoring. However, extracting breathing rate (BR) and heart rate (HR) from smartphone acoustic reflections remains challenging in indoor environments because thoracic reflections are weak and are often mixed with static clutter, hand motion, environmental multipath, and other dynamic sources. In this work, we present a smartphone-based frequency-modulated continuous wave (FMCW) acoustic sensing system that enables simultaneous BR and HR estimation using the integrated speaker and two physical microphones. Instead of processing the received signal as a single, mixed signal, the proposed method leverages distance information from the FMCW beat frequency and an angular phase index (AoA information), derived from dual-microphone and virtual aperture processing, to organize moving reflectors into a joint distance–angle–time representation. A 3D-DBSCAN clustering module is then applied to this representation to separate candidate dynamic sources from static and multipath components, without presupposing the number of sources. To further handle ambiguous cases where multiple candidate dynamic sources are detected, a Siamese similarity network is introduced as a conditional second-stage source-association module. The Siamese model compares candidate thoracic waveforms and estimates whether multiple detected components are likely to originate from the same physical source or different sources, thus improving source selection without resorting to classical blind source separation. The system was evaluated on 20 participants in two indoor environments, a laboratory and a bedroom, using three consumer smartphones and an electrocardiogram (ECG) reference device. In the smartphone-only blind configuration, the proposed pipeline achieved MAEs of 2.312 bpm for HR and 1.394 bpm for BR. In the ECG-assisted calibrated configuration, which is used to evaluate physiological coherence rather than deployable smartphone-only performance, the errors decreased to 0.462 bpm for HR and 0.091 bpm for BR. These results demonstrate that spatial clustering and conditional Siamese source pairing improve the robustness of acoustic vital sign detection using smartphones in indoor environments. Full article

Traditional methods for detecting the soluble solid content (SSC) of kumquats are often destructive, time-consuming, and inefficient. In this study, a multi-scale convolutional neural network (MS-CNN)-based method is proposed for the rapid and non-destructive prediction of kumquat SSC. By integrating near-infrared spectroscopy (900–1700 nm) with deep learning, 424 spectral samples of kumquats were collected and modeled using the MS-CNN framework. The proposed model adopts a multi-scale feature extraction structure inspired by the Inception architecture, which effectively enhances the representation of spectral features and reduces overfitting. Experimental results showed that the MS-CNN achieved an $R_{\mathrm{p}}^2$ of 0.88, an RMSEP of 0.62 °Brix, and an MAEP of 0.51 °Brix on the internal prediction set. Among the evaluated models, the MS-CNN achieved the highest $R_{\mathrm{p}}^2$, while its RMSEP was comparable to that of PLSR and lower than those of SVR, BP, CNN, and BiLSTM. The proposed approach enables fast, accurate, and non-destructive prediction of kumquat SSC, providing a novel technical solution for fruit quality assessment. This work holds significant theoretical and practical value, and future efforts will focus on expanding the dataset, optimizing the network structure, exploring multi-index joint prediction, and promoting its real-world application. Full article

Studying photovoltaics in engineering and science curricula is a time-consuming and expensive activity. This is why simulations are used, which, however, do not allow direct observation of the physical phenomena occurring in the devices. Based on an array of LEDs (light-emitting diodes) in photodetection mode, a low-cost educational platform for simulating photovoltaic systems including bypass and blocking diodes was developed. This allowed for the experimental characterization of the system’s I-V and P-V characteristics, obtained with a variable-load method under controlled lighting, as well as the qualitative reproduction of key photovoltaic phenomena such as mismatch and bypass diode activation. Additionally, the system allows for quantitative analyses starting from a reference value of 25 μW, obtained under full illumination conditions. This value will inevitably decrease as the platform’s operating conditions worsen, intentionally generated to study the behavior of the platform. Although the method does not provide a representation of the real photovoltaic field, it provides a simple and low-cost tool for the experimental study of photovoltaic behavior. The paper has been conceived for educational purposes, oriented towards laboratory teaching activities. Full article

Automated compliance checking for building fire protection design requires accurate extraction of domain-specific entities from technical code provisions. However, building fire protection codes contain highly specialized terminology, hierarchical clause structures, and complex regulatory semantics, which make manual information extraction time-consuming and error-prone. This study develops a domain-specific named entity recognition approach for Chinese building fire protection design codes. An expert-annotated dataset was constructed from five representative codes, containing 2748 annotated provisions and 13,877 entity mentions across nine entity categories. A RoBERTa-BiLSTM-CRF model was then developed to capture contextual semantic representations, bidirectional sequence dependencies, and label-transition constraints. Results suggest that the proposed model achieves 88.34%, 88.31%, and 88.32% for precision, recall, and F1-score, respectively. The extracted entities can be organized into structured and traceable records, providing a foundation for downstream building fire protection knowledge management and automated compliance checking. Full article