Search Results (20)

Industry 5.0 emphasizes human-centered integration of artificial intelligence in industrial contexts, yet successful adoption depends critically on workforce perception and acceptance. This research develops and validates a machine learning framework for predicting AI-related perceptions and expected impacts in the construction industry under small sample constraints typical of specialized industrial surveys. Specifically, the study aims to develop and empirically validate a predictive AI decision support model that estimates the expected impact of AI adoption in the construction sector based on digital competencies, ICT utilization, AI training and experience, and AI usage at both individual and organizational levels, operationalized through a composite AI Impact Index and two process-oriented outcomes (perceived task automation and perceived cost reduction). Using a dataset of 51 survey responses from Slovak construction professionals collected in 2025, we implement a methodologically rigorous approach specifically designed for limited-data regimes. The framework encompasses ordinal target simplification from five to three classes, dimensionality reduction through theoretically grounded composite indices reducing features from 15 to 7, exclusive deployment of low variance regularized models, and leave-one-out cross-validation for unbiased performance estimation. The optimal model (Lasso regression with recursive feature elimination) predicts cost reduction perception with R² = 0.501, MAE = 0.551, and RMSE = 0.709, while six classification targets achieve weighted F1 = 0.681, representing statistically optimal performance given sample constraints and perception measurement variability. Comparative evaluation confirms regularized models outperform high variance alternatives: random forest (R² = 0.412) and gradient boosting (R² = 0.292) exhibit substantially lower generalization performance, empirically validating the bias-variance trade-off rationale. Key methodological contributions include explicit bias-variance optimization preventing overfitting, feature selection via RFE reducing input space to six predictors (personal AI usage, AI impact on budgeting, ICT utilization, AI training, company size, and age), and demonstration that principled statistical approaches achieve meaningful predictions without requiring large-scale datasets or complex architectures. The framework provides a replicable blueprint for perception and impact prediction in data-constrained Industry 5.0 contexts, enabling targeted interventions, including customized training programs, strategic communication prioritization, and resource allocation for change management initiatives aligned with predicted adoption patterns. Full article

Background: Sickle cell disease (SCD) is a prevalent hereditary hemoglobinopathy associated with substantial morbidity, particularly in regions with limited access to advanced laboratory diagnostics. Conventional diagnostic workflows, including manual peripheral blood smear examination and biochemical or molecular assays, are resource-intensive, time-consuming, and subject to observer variability. Recent advances in artificial intelligence (AI) enable automated analysis of blood smear images and offer a scalable alternative for SCD screening. Methods: This study presents a controlled benchmark of CNNs, Vision Transformers, hierarchical Transformers, and hybrid CNN–Transformer architectures for image-level SCD classification using a publicly available peripheral blood smear dataset. Eleven ImageNet-pretrained models were fine-tuned under identical conditions using an explicit leakage-safe evaluation protocol, incorporating duplicate-aware, group-based data splitting and repeated splits to assess robustness. Performance was evaluated using accuracy and macro-averaged precision, recall, and F1-score, complemented by bootstrap confidence intervals, paired statistical testing, error-type analysis, and explainable AI (XAI). Results: Across repeated group-aware splits, CNN-based and hybrid architectures demonstrated more stable and consistently higher performance than transformer-only models. MaxViT-Tiny and DenseNet121 ranked highest overall, while pure ViTs showed reduced effectiveness under data-constrained conditions. Error analysis revealed a dominance of false-positive predictions, reflecting intrinsic morphological ambiguity in challenging samples. XAI visualizations suggest that CNNs focus on localized red blood cell morphology, whereas hybrid models integrate both local and contextual cues. Conclusions: Under limited-data conditions, convolutional inductive bias remains critical for robust blood-smear-based SCD classification. CNN and hybrid CNN–Transformer models offer interpretable and reliable performance, supporting their potential role as decision-support tools in screening-oriented research settings. Full article

Fingerprint recognition systems have relied on fragile workflows based on minutiae extraction, which suffer from significant performance losses under real-world conditions such as sensor diversity and low image quality. This study introduces a fully minutiae-free fingerprint recognition framework based on self-supervised Vision Transformers. A systematic evaluation of multiple DINOv2 model variants is conducted, and the proposed system ultimately adopts the DINOv2-Base Vision Transformer as the primary configuration, as it offers the best generalization performance trade-off under conditions of limited fingerprint data. Larger variants are additionally analyzed to assess scalability and capacity limits. The DINOv2 pretrained network is fine-tuned using self-supervised domain adaptation on 64,801 fingerprint images, eliminating all classical enhancement, binarization, and minutiae extraction steps. Unlike the single-sensor protocols commonly used in the literature, the proposed approach is extensively evaluated in a heterogeneous testbed with a wide range of sensors, qualities, and acquisition methods, including 1631 unique fingers from 12 datasets. The achieved EER of 5.56% under these challenging conditions demonstrates clear cross-sensor superiority over traditional systems such as VeriFinger (26.90%) and SourceAFIS (41.95%) on the same testbed. A systematic comparison of different model capacities shows that moderate-scale ViT models provide optimal generalization under limited-data conditions. Explainability analyses indicate that the attention maps of the model trained without any minutiae information exhibit meaningful overlap with classical structural regions (IoU = 0.41 ± 0.07). Openly sharing the full implementation and evaluation infrastructure makes the study reproducible and provides a standardized benchmark for future research. Full article

Predicting the fatigue lifespan of Twisted String Actuators (TSAs) is essential for improving the reliability of robotic and mechanical systems that rely on flexible transmission mechanisms. Traditional empirical approaches based on regression or Weibull distribution analysis have provided useful approximations, yet they often struggle to capture nonlinear dependencies and stochastic influences inherent to real-world fatigue behavior. This study introduces and compares four machine learning (ML) models—Linear Regression, Random Forest, XGBoost, and Gaussian Process Regression (GPR)—for predicting TSA lifespan under varying weight (W), number of strings (N), and diameter (D) conditions. Building upon this comparison, a hybrid physics-guided model is proposed by integrating an empirical fatigue life equation with an XGBoost residual-correction model. Experimental data collected from repetitive actuation tests (144 valid samples) served as the basis for training and validation. The hybrid model achieved an R² = 0.9856, RMSE = 5299.47 cycles, and MAE = 3329.67 cycles, outperforming standalone ML models in cross-validation consistency (CV R² = 0.9752). The results demonstrate that physics-informed learning yields superior interpretability and generalization even in limited-data regimes. These findings highlight the potential of hybrid empirical–ML modeling for component life prediction in robotic actuation systems, where experimental fatigue data are scarce and operating conditions vary. Full article

This paper presents an integrated three-dimensional (3D) quality inspection system for mold manufacturing that addresses critical industrial constraints, including zero-shot generalization without retraining, complete decision traceability for regulatory compliance, and robustness under severe data shortages (<2% defect rate). Dual optical sensors (Photoneo MotionCam 3D and SICK Ruler) are integrated via affine transformation-based registration, followed by computer-aided design (CAD)-based classification using geometric feature matching to CAD specifications. Unsupervised defect detection combines density-based spatial clustering of applications with noise (DBSCAN) clustering, curvature analysis, and alpha shape boundary estimation to identify surface anomalies without labeled training data. Industrial validation on 38 product classes (3000 samples) yielded 99.00% classification accuracy and 99.12% macroscopic precision, outperforming Point-MAE (93.24%) trained under the same limited-data conditions. The CAD-based architecture enables immediate deployment via CAD reference registration, eliminating the five-day retraining cycle required for deep learning, essential for agile manufacturing. Processing time stability (0.47 s compared to 43.68 s for Point-MAE) ensures predictable production throughput. Defect detection achieved 98.00% accuracy on a synthetic validation dataset (scratches: 97.25% F1; dents: 98.15% F1). Full article

The Terracotta Army, an integral part of China’s cultural heritage, has suffered physical erosion like cracks and notches over time. Manual inspection methods are inefficient and subjective. This study proposes an automated defect detection system based on computer vision to enhance the efficiency and precision of detecting these defects. The system includes the following core modules: (1) high-resolution image acquisition, which ensures comprehensive and detailed data capture; (2) sophisticated image illumination processing, which compensates for varying lighting conditions and improves image quality; (3) advanced image data augmentation techniques, which enrich the dataset and improve the generalization ability of the detection model; and (4) accurate defect detection, which leverages state-of-the-art algorithms. In the experimental phase, the efficacy of the proposed approach was evaluated. Illumination-enhanced low-light images were used for data augmentation, and the generated images showed high similarity to the original images, as measured by PSNR and SSIM. The YOLOv10 algorithm was employed for defect detection and achieved average detection rates of 91.71% for cracks and 93.04% for abrasions. This research provides a scientific and efficient solution for cultural relic protection and offers a valuable reference for future research in heritage conservation. Full article

This study presents a comparative evaluation of traditional machine-learning (ML) and deep-learning (DL) approaches for detecting and counting Varroa destructor mites in hyperspectral images. As Varroa infestations pose a serious threat to honeybee health, accurate and efficient detection methods are essential. The ML pipeline—based on Principal Component Analysis (PCA), k-Nearest Neighbors (kNN), and Support Vector Machine (SVM)—was previously published and achieved high performance (precision = 0.9983, recall = 0.9947), with training and inference completed in seconds on standard CPU hardware. In contrast, the DL approach, employing Faster R-CNN with ResNet-50 and ResNet-101 backbones, was fine-tuned on the same manually annotated images. Despite requiring GPU acceleration, longer training times, and presenting a reproducibility challenges, the deep-learning models achieved precision of 0.966 and 0.971, recall of 0.757 and 0.829, and F1-Score of 0.848 and 0.894 for ResNet-50 and ResNet-101, respectively. Qualitative results further demonstrate the robustness of the ML method under limited-data conditions. These findings highlight the differences between ML and DL approaches in resource-constrained scenarios and offer practical guidance for selecting suitable detection strategies. Full article

The ship waterjet propulsion system is a crucial power unit for high-performance vessels, and the operational state of its core component, the waterjet pump, is directly related to navigation safety and mission reliability. To enhance the intelligence and accuracy of pump fault diagnosis, this paper proposes a novel diagnostic framework that integrates a supervised autoencoder (SAE) with a large language model (LLM). This framework first employs an SAE to perform task-oriented feature learning on raw vibration signals collected from the pump’s guide vane casing. By jointly optimizing reconstruction and classification losses, the SAE extracts deep features that both represent the original signal information and exhibit high discriminability for different fault classes. Subsequently, the extracted feature vectors are converted into text sequences and fed into an LLM. Leveraging the powerful sequential information processing and generalization capabilities of LLM, end-to-end fault classification is achieved through parameter-efficient fine-tuning. This approach aims to avoid the traditional dependence on manually extracted time-domain and frequency-domain features, instead guiding the feature extraction process via supervised learning to make it more task-specific. To validate the effectiveness of the proposed method, we compare it with a baseline approach that uses manually extracted features. In two experimental scenarios, direct diagnosis with full data and transfer diagnosis under limited-data, cross-condition settings, the proposed method significantly outperforms the baseline in diagnostic accuracy. It demonstrates excellent performance in automated feature extraction, diagnostic precision, and small-sample data adaptability, offering new insights for the application of large-model techniques in critical equipment health management. Full article

Fault diagnosis accuracy in lithium-ion battery-based energy storage systems is significantly constrained by the limited availability of fault-specific datasets. This study addresses this critical issue by proposing a diffusion-based data augmentation methodology tailored explicitly for battery fault diagnosis scenarios. The proposed conditional diffusion model leverages transfer learning and attention-enhanced fine-tuning strategies to generate high-quality synthetic fault data, ensuring targeted representation of rare fault conditions. By integrating condition-aware sampling strategies, the approach effectively mitigates mode collapse issues frequently encountered in adversarial generative methods, thus substantially enriching the diversity and quality of fault representations. Comprehensive evaluation using statistical similarity measures and downstream classification tasks demonstrates notable improvements. After the integration of attention mechanisms, the Pearson correlation coefficient between the synthetic and real samples increases from 0.29 to 0.91. In downstream diagnostic tasks, models trained on augmented datasets exhibit substantial gains in regards to the recall and F1-score, which increase from near-zero levels to values exceeding 0.91 for subtle overcharge and overdischarge faults. These results confirm the effectiveness and practical utility of the proposed augmentation approach in enhancing diagnostic performance under data-scarce conditions. Full article

To address the issue of insufficient prediction accuracy in traditional GM(1,1) models caused by significant nonlinear fluctuations in time-series data for ancient building structural health monitoring, this study proposes a wavelet decomposition-based GM(1,1)-BP neural network coupled prediction model. By constructing a multi-scale fusion framework, we systematically resolve the collaborative optimization between trend prediction and detail modeling. The methodology comprises four main phases: First, wavelet transform is employed to decompose original monitoring sequences into time-frequency components, obtaining low-frequency trends characterizing long-term deformation patterns and high-frequency details reflecting dynamic fluctuations. Second, GM(1,1) models are established for the trend extrapolation of low-frequency components, capitalizing on their advantages in limited-data modeling. Subsequently, BP neural networks are designed for the nonlinear mapping of high-frequency components, leveraging adaptive learning mechanisms to capture detail features induced by environmental disturbances and complex factors. Finally, a wavelet reconstruction fusion algorithm is developed to achieve the collaborative optimization of dual-channel prediction results. The model innovatively introduces a detail information correction mechanism that simultaneously overcomes the limitations of single grey models in modeling nonlinear fluctuations and enhances neural networks’ capability in capturing long-term trend features. Experimental validation demonstrates that the fused model reduces the Root Mean Square Error (RMSE) by 76.5% and 82.6% compared to traditional GM(1,1) and BP models, respectively, with the accuracy grade improving from level IV to level I. This achievement provides a multi-scale analytical approach for the quantitative interpretation of settlement deformation patterns in ancient architecture. The established “decomposition-prediction-fusion” technical framework holds significant application value for the preventive conservation of historical buildings. Full article

Understanding fish habitats is essential for fisheries management, habitat restoration, and species protection. Automated fish detection is a key tool in these applications, which enables real-time monitoring and quantitative analysis. Recent advancements in high-resolution cameras and machine learning technologies have facilitated image analysis automation, promoting remote fish tracking. However, many of these detection methods require large volumes of annotated data, which involve considerable effort and time. Additionally, their practical implementation remains challenging in environments with limited data. Hence, this study proposes an anomaly-based fish detection approach by integrating Patch Distribution Modeling with data augmentation techniques, including Affine Transformations, Neural Filters, and SinGAN. Field experiments were conducted in Lake Izunuma-Uchinuma, Japan, using an electrofishing boat to acquire data. Evaluation metrics, such as AUROC and F1-score, assessed detection performance. The results indicate that, compared to the original dataset (AUROC: 0.836, F1-score: 0.483), Neural Filters (AUROC: 0.940, F1-score: 0.879) and Affine Transformations (AUROC: 0.942, F1-score: 0.766) improve anomaly detection. However, SinGAN exhibited no measurable enhancement, indicating the necessity for further optimization. This shows the potential of the proposed approach to enhance automated fish detection in limited-data environments, supporting aquatic ecosystem sustainability. Full article

Conventional strategies aimed at mitigating beam-hardening artifacts in computed tomography (CT) can be categorized into two main approaches: (1) postprocessing following conventional reconstruction and (2) iterative reconstruction incorporating a beam-hardening model. While the former fails in low-dose and/or limited-data cases, the latter substantially increases computational cost. Although deep learning-based methods have been proposed for several cases of limited-data CT, few works in the literature have dealt with beam-hardening artifacts, and none have addressed the problems caused by randomly selected projections and a highly limited span. We propose the deep learning-based prior image constrained (PICDL) framework, a hybrid method used to yield CT images free from beam-hardening artifacts in different limited-data scenarios based on the combination of a modified version of the Prior Image Constrained Compressed Sensing (PICCS) algorithm that incorporates the L2 norm (L2-PICCS) with a prior image generated from a preliminary FDK reconstruction with a deep learning (DL) algorithm. The model is based on a modification of the U-Net architecture, incorporating ResNet-34 as a replacement of the original encoder. Evaluation with rodent head studies in a small-animal CT scanner showed that the proposed method was able to correct beam-hardening artifacts, recover patient contours, and compensate streak and deformation artifacts in scenarios with a limited span and a limited number of projections randomly selected. Hallucinations present in the prior image caused by the deep learning model were eliminated, while the target information was effectively recovered by the L2-PICCS algorithm. Full article

With increasing wafer sizes and diversifying die patterns, automated optical inspection (AOI) is progressively replacing traditional visual inspection (VI) for wafer defect detection. Yet, the defect classification efficacy of current AOI systems in our case company is not optimal. This limitation is due to the algorithms’ reliance on expertly designed features, reducing adaptability across various product models. Additionally, the limited time available for operators to annotate defect samples restricts learning potential. Our study introduces a novel hybrid self-training algorithm, leveraging semi-supervised learning that integrates pseudo-labeling, noisy student, curriculum labeling, and the Taguchi method. This approach enables classifiers to autonomously integrate information from unlabeled data, bypassing the need for feature extraction, even with scarcely labeled data. Our experiments on a small-scale set show that with 25% and 50% labeled data, the method achieves over 92% accuracy. Remarkably, with only 10% labeled data, our hybrid method surpasses the supervised DenseNet classifier by over 20%, achieving more than 82% accuracy. On a large-scale set, the hybrid method consistently outperforms other approaches, achieving up to 88.75%, 86.31%, and 83.61% accuracy with 50%, 25%, and 10% labeled data. Further experiments confirm our method’s consistent superiority, highlighting its potential for high classification accuracy in limited-data scenarios. Full article

Training a model to recognize human actions in videos is computationally intensive. While modern strategies employ transfer learning methods to make the process more efficient, they still face challenges regarding flexibility and efficiency. Existing solutions are limited in functionality and rely heavily on pretrained architectures, which can restrict their applicability to diverse scenarios. Our work explores knowledge distillation (KD) for enhancing the training of self-supervised video models in three aspects: improving classification accuracy, accelerating model convergence, and increasing model flexibility under regular and limited-data scenarios. We tested our method on the UCF101 dataset using differently balanced proportions: 100%, 50%, 25%, and 2%. We found that using knowledge distillation to guide the model’s training outperforms traditional training without affecting the classification accuracy and while reducing the convergence rate of model training in standard settings and a data-scarce environment. Additionally, knowledge distillation enables cross-architecture flexibility, allowing model customization for various applications: from resource-limited to high-performance scenarios. Full article

In industrial quality control, especially in the field of manufacturing defect detection, deep learning plays an increasingly critical role. However, the efficacy of these advanced models is often hindered by their need for large-scale, annotated datasets. Moreover, these datasets are mainly based on RGB images, which are very different from X-ray images. Addressing this limitation, our research proposes a methodology that incorporates domain-specific self-supervised pretraining techniques using X-ray imaging to improve defect detection capabilities in manufacturing products. We employ two pretraining approaches, SimSiam and SimMIM, to refine feature extraction from manufacturing images. The pretraining stage is carried out using an industrial dataset of 27,901 unlabeled X-ray images from a manufacturing production line. We analyze the performance of the pretraining against transfer-learning-based methods in a complex defect detection scenario using a Faster R-CNN model. We conduct evaluations on both a proprietary industrial dataset and the publicly available GDXray dataset. The findings reveal that models pretrained with domain-specific X-ray images consistently outperform those initialized with ImageNet weights. Notably, Swin Transformer models show superior results in scenarios rich in labeled data, whereas CNN backbones are more effective in limited-data environments. Moreover, we underscore the enhanced ability of the models pretrained with X-ray images in detecting critical defects, crucial for ensuring safety in industrial settings. Our study offers substantial evidence of the benefits of self-supervised learning in manufacturing defect detection, providing a solid foundation for further research and practical applications in industrial quality control. Full article