Search Results (89)

Digital image quality is crucial for reliable analysis in applications such as medical imaging, satellite remote sensing, and video surveillance. However, traditional denoising methods struggle to balance noise removal with detail preservation and lack adaptability to various types of noise. We propose a novel three-module architecture for image denoising, comprising a generator, a dual-path-UNet-based denoiser, and a discriminator. The generator creates synthetic noise patterns to augment training data, while the dual-path-UNet denoiser uses multiple receptive field modules to preserve fine details and dense feature fusion to maintain global structural integrity. The discriminator provides adversarial feedback to enhance denoising performance. This dual-path adversarial training mechanism addresses the limitations of traditional methods by simultaneously capturing both local details and global structures. Experiments on the SIDD, DND, and PolyU datasets demonstrate superior performance. We compare our architecture with the latest state-of-the-art GAN variants through comprehensive qualitative and quantitative evaluations. These results confirm the effectiveness of noise removal with minimal loss of critical image details. The proposed architecture enhances image denoising capabilities in complex noise scenarios, providing a robust solution for applications that require high image fidelity. By enhancing adaptability to various types of noise while maintaining structural integrity, this method provides a versatile tool for image processing tasks that require preserving detail. Full article

The rapid proliferation of wearable sensors and advanced tracking technologies has revolutionized data collection in elite sports, enabling continuous monitoring of athletes’ physiological and biomechanical states. This study proposes a comprehensive big data analytics framework that integrates data acquisition, processing, analytics, and decision support, demonstrated through synthetic datasets in football, basketball, and athletics case scenarios, modeled to represent typical data patterns and decision-making workflows observed in elite sport environments. Analytical methods, including gradient boosting classifiers, logistic regression, and multilayer perceptron models, were employed to predict injury risk, optimize in-game tactical decisions, and personalize sprint mechanics training. Key results include a 12% reduction in hamstring injury rates in football, a 16% improvement in clutch decision-making accuracy in basketball, and an 8% decrease in 100 m sprint times among athletes. The framework’s visualization tools and alert systems supported actionable insights for coaches and medical staff. Challenges such as data quality, privacy compliance, and model interpretability are addressed, with future research focusing on edge computing, federated learning, and augmented reality integration for enhanced real-time feedback. This study demonstrates the potential of integrated big data analytics to transform sports performance optimization, offering a reproducible and ethically sound platform for advancing personalized, data-driven athlete management. Full article

This paper presents an initial output of the project “Augmented Intelligence in Mathematics Education through Modeling, Automatic Reasoning and Artificial Intelligence (IAxEM-CM/PHS-2024/PH-HUM-383)”. The starting hypothesis of this project is that the use of technological tools, such as mathematical modeling, visualization, automatic reasoning and artificial intelligence, significantly improves the teaching and learning of mathematics, in addition to fostering positive attitudes in students. With this hypothesis in mind, in this article, we describe an investigation that has been developed in initial training courses for mathematics teachers in several universities in Madrid, where students used GeoGebra Discovery automated reasoning tools to explore geometric properties in real objects through mathematical paths. Through these activities, future teachers modeled, conjectured and validated geometric relationships directly on photographs of their environment, with the essential concourse of the automated discovery and verification of geometric properties provided by GeoGebra Discovery. The feedback provided by the students’ answers to a questionnaire concerning this novel approach shows a positive evaluation of the experience, especially in terms of content learning and the practical use of technology. Although technological, pedagogical and disciplinary knowledge is well represented, the full integration of these components (according to the TPACK model) is still incipient. Finally, the formative potential of the approach behind this experience is highlighted in a context where Artificial Intelligence tools have an increasing presence in education, as well as the need to deepen these three kinds of knowledge in similar experiences that articulate them in a more integrated way. Full article

Craniofacial deformities, particularly craniosynostosis, present significant surgical challenges due to complex anatomy and the need for individualised, high-precision interventions. Augmented reality (AR) has emerged as a promising tool in craniofacial surgery, offering enhanced spatial visualisation, real-time anatomical referencing, and improved surgical accuracy. This review explores the current and emerging applications of AR in preoperative planning, intraoperative navigation, and surgical education within paediatric craniofacial surgery. Through a literature review of peer-reviewed studies, we examine how AR platforms, such as the VOSTARS system and Microsoft HoloLens, facilitate virtual simulations, precise osteotomies, and collaborative remote guidance. Despite demonstrated benefits in feasibility and accuracy, widespread clinical adoption is limited by technical, ergonomic, financial, and training-related challenges. Future directions include the integration of artificial intelligence, haptic feedback, and robotic assistance to further augment surgical precision and training efficacy. AR holds transformative potential for improving outcomes and efficiency in craniofacial deformity correction, warranting continued research and clinical validation. Full article

This study aims to identify and evaluate the essential design features, strengths, and limitations of a virtual reality (VR) application that has been developed to train an international sales force effectively for a premium global wine brand. The study emphasizes the value of stakeholder-driven iterative development and systematic evaluations. A case study methodology was adopted for the research, focusing on a VR training application, developed for Marchesi Antinori. The Scrum framework was employed to facilitate iterative stakeholder collaboration. A qualitative evaluation was conducted using focus groups, comprising marketing, communications, and sales representatives. A systematic application of natural language processing (NLP) embedding techniques and recursive clustering analyses was undertaken to interpret stakeholder feedback. The findings suggest that stakeholder-driven, iterative processes can significantly enhance the effectiveness of VR applications by providing a clear structure for immersive storytelling that focuses on terroir characteristics, vineyard operations, and cellar practices. Stakeholders acknowledged the potent educational benefits of VR in regard to business-to-business (B2B) sales training. However, they also highlighted significant limitations, including user discomfort, concerns about authenticity, and variations in market receptivity. Alternative immersive technologies, including augmented reality and immersive multimedia environments, have emerged as valuable complementary approaches. This study addresses a significant gap in the literature by examining the application of VR technology for B2B sales training in the premium wine industry. The study integrates an iterative Scrum methodology with advanced natural language processing (NLP) analytical techniques to derive nuanced, context-rich insights. Full article

The growing complexity and risk profile of fire and emergency incidents necessitate advanced training methodologies that go beyond traditional approaches. Live-fire drills and classroom-based instruction, while foundational, often fall short in providing safe, repeatable, and scalable training environments that accurately reflect the dynamic nature of real-world emergencies. Recent advancements in immersive technologies, including virtual reality (VR), augmented reality (AR), mixed reality (MR), extended reality (XR), and simulation-based systems, offer promising alternatives to address these challenges. This review provides a comprehensive overview of the integration of VR, AR, MR, XR, and simulation technologies into firefighter and incident commander training. It examines current practices across fire services and emergency response agencies, highlighting the capabilities of immersive and interactive platforms to enhance operational readiness, decision-making, situational awareness, and team coordination. This paper analyzes the benefits of these technologies, such as increased safety, cost-efficiency, data-driven performance assessment, and personalized learning pathways, while also identifying persistent challenges, including technological limitations, realism gaps, and cultural barriers to adoption. Emerging trends, such as AI-enhanced scenario generation, biometric feedback integration, and cloud-based collaborative environments, are discussed as future directions that may further revolutionize fire service education. This review aims to support researchers, training developers, and emergency service stakeholders in understanding the evolving landscape of digital training solutions, with the goal of fostering more resilient, adaptive, and effective emergency response systems. Full article

This study evaluates the effectiveness of augmented reality (AR), using the Microsoft™ HoloLens™™ 2, for identifying and localizing PCB components compared to traditional PDF-based methods. Two experiments examined the influence of user expertise, viewing angles, and component sizes on accuracy and usability. The results indicate that AR improved identification accuracy and user experience for non-experts, although it was slower than traditional methods for experienced users. Optimal performance was achieved at 90° viewing angles, while accuracy declined significantly at oblique angles. Medium-sized components received the highest confidence scores, suggesting favorable visibility and recognition characteristics within this group, though further evaluation with a broader component distribution is warranted. Participant feedback highlighted the system’s intuitive interface and effective guidance, but also noted challenges with marker stability, visual discomfort, and ergonomic limitations. These findings suggest that AR can enhance training and reduce errors in electronics manufacturing, although refinements in marker rendering and user onboarding are necessary to support broader adoption. This research provides empirical evidence on the role of AR in supporting user-centered design and improving task performance in industrial electronics workflows. Full article

This work proposes a hybrid deep learning-based framework for visual feedback control in an eye-in-hand robotic system. The framework uses an early fusion approach in which real and synthetic images define the training data. The first layer of a ResNet-18 backbone is augmented to fuse interest-point maps with RGB channels, enabling the network to capture scene geometry better. A manipulator robot with an eye-in-hand configuration provides a reference image, while subsequent poses and images are generated synthetically, removing the need for extensive real data collection. The experimental results reveal that this enriched input representation significantly improves convergence accuracy and velocity smoothness compared to a baseline that processes real images alone. Specifically, including feature point maps allows the network to discriminate crucial elements in the scene, resulting in more precise velocity commands and stable end-effector trajectories. Thus, integrating additional, synthetically generated map data into convolutional architectures can enhance the robustness and performance of the visual servoing system, particularly when real-world data gathering is challenging. Unlike existing visual servoing methods, our early fusion strategy integrates feature maps directly into the network’s initial convolutional layer, allowing the model to learn critical geometric details from the very first stage of training. This approach yields superior velocity predictions and smoother servoing compared to conventional frameworks. Full article

Wearable sensors are increasingly used to provide objective quantification of spatiotemporal and kinematic parameters post-stroke. This study aimed to evaluate the practical value of sensor-based gait reports in delivering augmented feedback and informing the development of home training programmes following a 2-week supervised intensive intervention after stroke. Four patients with chronic stroke were assessed on four occasions (pre- and post-intervention, 3-month, and 6-month follow-ups) using clinical gait tests, during which a portable sensor-based system recorded kinematic data. The meaningfulness of individual changes in gait parameters was interpreted based on established minimal detectable change values (MDC). Three participants improved their gait speed, joint angles, and/or cadence in the Ten-Metre Walk Test, and three participants improved their walking distance in the Six-Minute Walk Test. The improvements were most evident at the 3-month follow-up (with the most obvious changes above MDC estimates) and indicated the reappearance of normal gait patterns, adjustments of gait patterns, or a combination of both. Participants showed interest in and understanding of the information derived from the gait reports (ratings of 5–10 out of 10). In conclusion, augmented feedback derived from gait reports provides a valuable complement to traditional clinical assessments used in stroke rehabilitation to optimize treatment outcomes. Full article

AI-oriented quality inspection in manufacturing often faces highly imbalanced data, as defective products are rare, and there are limited possibilities for data augmentation. This paper presents a systematic comparison between Deep Transfer Learning (DTL) and Contrastive Learning (CL) under such challenging conditions, addressing a critical gap in the industrial machine learning literature. We focus on a galvanized steel coil quality classification task with acceptable vs. defective classes, where the vast majority of samples (>95%) are acceptable. We implement a DTL approach using strategically fine-tuned YOLOv8 models pre-trained on large-scale datasets, and a CL approach using a Siamese network with multi-reference design to learn robust similarity metrics for one-shot classification. Experiments employ k-fold cross-validation and a held-out gold-standard test set of coil images, with statistical validation through bootstrap resampling. Results demonstrate that DTL significantly outperforms CL, achieving higher overall accuracy (81.7% vs. 61.6%), F1-score (79.2% vs. 62.1%), and precision (91.3% vs. 61.0%) on the challenging test set. Computational analysis reveals that DTL requires 40% less training time and 25% fewer parameters while maintaining superior generalization capabilities. We provide concrete guidance on when to select DTL over CL based on dataset characteristics, demonstrating that DTL is particularly advantageous when data augmentation is constrained by domain-specific spatial patterns. Additionally, we introduce a novel adaptive inspection framework that integrates human-in-the-loop feedback with domain adaptation techniques for continuous model improvement in production environments. Our comprehensive comparative analysis offers empirically validated insights into performance trade-offs between these approaches under extreme class imbalance, providing valuable direction for practitioners implementing industrial quality inspection systems with limited, skewed datasets. Full article

Modern sport technology is a rapidly evolving scientific field utilizing leading techniques, such as the internet of things (IoT), augmented reality (AR), virtual reality (VR) and the use of many kinds of sensors. The feedback and the support provided by technology seem to have the potential to bring sufficient changes in sports training following a new trend, which is trying to integrate modern technology into sports. The aim of this study is to integrate responsive feedback and internet of things (IoT) technology into sports mainly in order to support the training of novice bowling athletes, raise their engagement with their sport and boost their performance. Following this direction, this research utilizes a system of sensors to apply step tracking of a bowling athlete during a throw ball attempt. The proposed solution is a European-funded project (ERASMUS+) supporting the “Europe 2020 strategy” to motivate people to transform their learning experiences into a beneficial way of constructing better performance. The results of the proposed system evaluation showed that bowlers and their coaches highlighted its usability and its usefulness. Thus, the proposed system may prove a valuable tool supporting athletes in general, and more particularly, novices’ training in bowling. Full article

Accurate acne grading is critical in optimizing therapeutic decisions yet remains challenging due to lesion ambiguity and subjective clinical assessments. This study proposes the Feature Feedback-Based Pseudo-Label Learning (FF-PLL) framework to address these limitations through three innovations: (1) an acne feature feedback (AFF) architecture with iterative pseudo-label refinement to improve the training robustness, enhance the pseudo-label quality, and increase the feature diversity; (2) all-facial skin segmentation (AFSS) to reduce background noise, enabling precise lesion feature extraction; and (3) the AcneAugment (AA) strategy to foster model generalization by introducing diverse acne lesion representations. Experiments on the ACNE04 and ACNE-ECKH benchmark datasets demonstrate the superiority of the proposed framework, achieving accuracy of 87.33% on ACNE04 and 67.50% on ACNE-ECKH. Additionally, the model attains sensitivity of 87.31%, specificity of 90.14%, and a Youden index (YI) of 77.45% on ACNE04. These advancements establish FF-PLL as a clinically viable solution for standardized acne assessment, bridging critical gaps between computational dermatology and practical healthcare needs. Full article

Robot-assisted minimally invasive surgery offers advantages over traditional laparoscopic surgery, including precision and improved patient outcomes. However, its complexity requires extensive training, leading to the development of simulators that still face challenges such as limited feedback and lack of realism. This study presents an augmented reality-based surgical simulator tailored for minimally invasive pancreatic surgery, integrating an innovative parallel robot, real-time AI-driven force estimation, and haptic feedback. Using Unity and the HoloLens 2, the simulator offers a realistic augmented environment, enhancing spatial awareness and planning in surgical scenarios. A convolutional neural network (CNN) model predicts forces without physical sensors, achieving a mean absolute error of 0.0244 N. Tests indicate a strong correlation between applied and predicted forces, with a haptic feedback latency of 65 ms, suitable for real-time applications. Its modularity makes the simulator accessible for training and preoperative planning, addressing gaps in current robotic surgery training tools while setting the stage for future improvements and broader integration. Full article

Background: Motor learning can occur through active reaching with the arm hidden from view, leading to improvements in somatosensory acuity and modulation of functional connectivity in sensorimotor and reward networks. In this proof-of-principle study, we assess if the same paradigm benefits stroke survivors using a compact end-effector robot with integrated gaming elements. Methods: Nine community-dwelling chronic hemiplegic stroke survivors with persistent somatosensory deficits participated in 15 training sessions, each lasting 1 h. Every session comprised a robotic-based joint approximation block, followed by 240 repetitions of training using a forward-reaching task with the affected forearm covered from view. During movement, the robot provided haptic guidance along the movement path as enhanced sensory cues. Augmented reward feedback was given following every successful movement as positive reinforcement. Baseline, post-intervention, and 1-month follow-up assessments were conducted, with the latter two sessions occurring after the final training day. Results: Training led to reliable improvements in endpoint accuracy, faster completion times, and smoother movements. Acceptability and feasibility analyses were performed to understand the viability of the intervention. Significant improvement was observed mainly in robotic-based sensory outcomes up to a month post training, suggesting that training effects were predominantly sensory, rather than motor. Conclusions: The study outcomes provide preliminary evidence supporting the feasibility of this intervention for future adoption in neurorehabilitation. Full article

Augmented reality (AR) technologies can improve the quality of life of individuals with visual impairments. The current study evaluated the efficacy of Retiplus, a new AR-based low-vision device, which was designed to enhance spatial awareness and visual function in patients with peripheral visual field loss. Thirteen patients diagnosed with retinitis pigmentosa (RP) participated in this study. The patients’ visual acuity, visual field, and subjective perception of peripheral vision and mobility were assessed both without and with the AR aid, following a training period consisting of five 1 h sessions. The results showed a significant expansion of the visual field (VF) in all four quadrants (right, left, upper, and lower) with a greater horizontal diameter enlargement (21.38° ± 12.94°) than vertical (15° ± 10.08°), with a statistically significant difference. However, the increase in VF was accompanied by a modest reduction in visual acuity due to the minification of the image on the display. Patient feedback also highlighted significant benefits on the ability to perform activities of daily living (ADL) in low-light environments and improved spatial orientation, suggesting that the AR system is helpful for some limitations imposed by patients’ conditions. These findings underscore the importance of optimizing AR technology to support visually impaired populations. Full article