Search Results (240)

This paper presents a real-time pedestrian situation detection framework for autonomous mobility platforms. The proposed approach extracts pedestrians from images acquired by a camera mounted on an autonomous mobility system, classifies their postures, tracks their trajectories, and subsequently detects pedestrian situations. A convolutional neural network (CNN) is employed for pedestrian detection and posture classification, where the YOLOv12 model is fine-tuned via transfer learning for this purpose. To improve detection and classification performance, a region of interest (ROI) is defined using camera calibration data, enabling robust detection of small-scale pedestrians over long distances. Using a custom-labeled dataset, the proposed method achieves a precision of 96.6% and a recall of 97.0% for pedestrian detection and posture classification. The detected pedestrians are tracked using the DeepSORT algorithm, and their situations are inferred through a rule-based analysis module. Experimental results demonstrate that the proposed system operates at an execution speed of 58.11 ms per frame, corresponding to 17.2 fps, thereby satisfying the real-time requirements for autonomous mobility applications. These results confirm that the proposed framework enables reliable real-time pedestrian extraction and situation awareness in real-world autonomous mobility environments. Full article

Academic engagement is a multidimensional construct encompassing students’ cognitive, emotional, and behavioral investment in learning. This study examines the levels and predictors of academic engagement among 390 students enrolled in the Master’s in Secondary Education Teacher Training at the National University of Distance Education (UNED, Spain). Using the Utrecht Work Engagement Scale-Student (UWES-S) and a quantitative, cross-sectional, and correlational design, the research explores associations between engagement and sociodemographic and motivational variables. Results indicate moderately high engagement levels, with dedication emerging as the most salient dimension, followed by absorption and vigor. Engagement correlated positively with age and was slightly higher among women, while vocational motivation stood out as the strongest differentiating factor. Prior teaching experience showed no significant influence. The findings highlight the importance of fostering purpose, professional meaning, and identity in initial teacher education—particularly in distance learning contexts—and suggest practical implications for designing supportive pedagogical environments that sustain students’ motivation and academic commitment. Full article

We propose a novel continual self-supervised learning (CSSL) framework for simultaneously learning diverse features from multi-window-obtained chest computed tomography (CT) images and ensuring data privacy. Achieving a robust and highly generalizable model in medical image diagnosis is challenging, mainly because of issues, such as the scarcity of large-scale, accurately annotated datasets and domain shifts inherent to dynamic healthcare environments. Specifically, in chest CT, these domain shifts often arise from differences in window settings, which are optimized for distinct clinical purposes. Previous CSSL frameworks often mitigated domain shift by reusing past data, a typically impractical approach owing to privacy constraints. Our approach addresses these challenges by effectively capturing the relationship between previously learned knowledge and new information across different training stages through continual pretraining on unlabeled images. Specifically, by incorporating a latent replay-based mechanism into CSSL, our method mitigates catastrophic forgetting due to domain shifts during continual pretraining while ensuring data privacy. Additionally, we introduce a feature distillation technique that integrates Wasserstein distance-based knowledge distillation and batch-knowledge ensemble, enhancing the ability of the model to learn meaningful, domain-shift-robust representations. Finally, we validate our approach using chest CT images obtained across two different window settings, demonstrating superior performance compared with other approaches. Full article

This paper explores the capability of Large Language Models (LLMs) to perform zero-shot planning through multimodal reasoning, with a particular emphasis on applications to Unmanned Ground Vehicles (UGVs) and unmanned platforms in general. We present a modular system architecture that integrates a general-purpose LLM with visual and spatial inputs for adaptive planning to iteratively guide UGV behavior. Although the framework is demonstrated in a ground-based setting, it directly extends to other unmanned systems, where semantic reasoning and adaptive planning are increasingly critical for autonomous mission execution. To assess performance, we employ a continuous evaluation metric that jointly considers distance and orientation, offering a more informative and fine-grained alternative to binary success measures. We evaluate a foundational LLM (i.e., Gemini 2.0 Flash, Google DeepMind) on a suite of zero-shot navigation and exploration tasks in simulated environments. Unlike prior LLM-robot systems that rely on fine-tuning or learned waypoint policies, we evaluate a purely zero-shot, stepwise LLM planner that receives no task demonstrations and reasons only from the sensed data. Our findings show that LLMs exhibit encouraging signs of goal-directed spatial planning and partial task completion, even in a zero-shot setting. However, inconsistencies in plan generation across models highlight the need for task-specific adaptation or fine-tuning. These findings highlight the potential of LLM-based multimodal reasoning to enhance autonomy in UGV and drone navigation, bridging high-level semantic understanding with robust spatial planning. Full article

This work aimed to develop and evaluate a real-time communication channel detection system in the Very High Frequency (VHF) band using software-defined radio (SDR). For this purpose, an FFT based spectral analyzer with 32,768 points was designed, capable of converting signals from the time domain to the frequency domain, ensuring efficient characterization of a 3 MHz bandwidth with updates every 0.5 s. Three detection algorithms were developed and compared: Energy Detection (ED) and two Waveform-Based Detection methods supported by machine learning models, SVM and KNN. ED stood out for its low computational requirements, suitable for low-cost systems, but had a limited probability of detection ($P_d$) at short distances, with zero detection beyond 500 m. KNN showed superior performance at longer distances, achieving 23% $P_d$ at 700 m but insufficient for real-time applications. The SVM model proved to be the most effective, achieving a $P_d$ of 80% at 1000 m and maintaining a low false positive rate of around 1%. It is concluded that the SVM model is the most suitable for real-time detection systems in the VHF band, offering a balance between accuracy and usability. The extrapolation of the results demonstrates the system’s potential for coverage greater than 2 km with higher-powered marine radios, around 25 W. Full article

This study examines the key factors influencing the quality of distance learning in higher education during the COVID-19 pandemic, a period when online learning became the dominant mode of education. Using a descriptive method and a 26-item questionnaire, data were collected from a representative sample of 360 students in Vojvodina, Serbia. The factors analyzed include computer literacy and technology access (Ph1), students’ ability to balance life obligations with study demands (Ph2), and their motivation for distance learning (Ph3). The results show that 89% of students had adequate IT access, 47% were able to reconcile study and personal obligations, and 70% reported strong motivation. Correlation analysis confirmed a statistically significant positive relationship between all three factors and students’ perceptions of well-organized distance learning, thus supporting the main research hypothesis. Beyond these findings, this study interprets digital literacy as adaptability, time management as resilience, and motivation as value orientation and future thinking—core dimensions of sustainability competences outlined in the European GreenComp framework. Distance learning is therefore positioned not only as an emergency response but also as a transformative pedagogy that integrates brain (knowledge), hands (skills), heart (values), and spirit (purpose), contributing to sustainable and resilient higher education. Full article

Background: The Derotational femoral osteotomy (DFO) is an effective surgical treatment for patients with disabling anterior knee pain associated with pathological Femoral anteversion (FAV). However, the complexity in determining the precise origin of the deformity has put limits on its use. This study aims to review the literature to learn how the authors study the origin of the deformity and then provide a new methodology using 3D technology to assess the origin of FAV. Methods: A search of the literature was conducted on PubMed utilizing the following search string: “anteversion” and “femur” or “origin” or “CT” or “MRI” or “3D”. In addition, an observational study was conducted on CT scans of six femurs from three female patients with unilateral pathological FAV. This work represents a pilot study and should be considered preliminary. Using the 3DSlicer (version 4.11.20210226), MeshMixer (version3.5), and 3DBuilder software (Microsoft.com), 3D biomodels were generated. A mirrored healthy femur served as a reference. The CloudCompare software (software version 2.13.0) was used to compare volumetric structures and analyze torsional deformities. Torsion at each level was quantified using MATLAB (software version 23.2). Results: The 3D technology identified three torsional patterns: 1. FAV predominantly originating at the femoral head (distance between the centroids = maximum deformity in the last discs, which coincides with the proximal region of the femur; heat maps = red in the proximal femur); 2. FAV primarily affects the mid-distal diaphysis (distance between the centroids = maximum deformity in the first discs, which coincides with mid-distal third of the femur; heat maps = red in the diaphyseal level); 3. a pan-diaphyseal deformity involving the entire femur (distance between the centroids = both the first and last discs, means deformity along the entire femur; heat maps = red along the entire femoral diaphysis). Conclusions: All femoral segments contributed to the total FAV, but the location and severity varied among the cases. Pathological FAV is a multifactorial deformity that can arise in different femoral regions. Individualized correction strategies are essential to improving DFO outcomes and preventing secondary deformities. It is important to note that the pilot data is intended to be purely illustrative and, as such, should not be utilized for the purposes of guiding clinical decision-making. Full article

Comparative investigations on homogenous cranial series have demonstrated that metopism is linked to a specific configuration of the cranial vault; however, there are no comparative data concerning the endocranial morphology in this condition. This study aimed to compare the endocranial space in metopic and control crania using morphometric analysis and machine learning algorithms. For this purpose, a series of 230 (184 control and 46 metopic) dry crania of contemporary adult Bulgarian males were scanned using an industrial µCT system. The 3D coordinates of 47 landmarks were collected on the endocranial surface. All possible measurements between the landmarks were calculated as Euclidean distances. The resultant 1081 measurements represented the initial dataset, which was reduced to smaller datasets applying different criteria. The derived datasets were used for learning a set of classification models by machine learning algorithms. The morphometric analysis showed that in the metopic crania some segments of the anterior and middle cranial fossae were significantly longer, and the landmark endobregma was significantly closer to the anterior and middle sections of the cranial base. The most accurate model, with a classification accuracy of 85%, was the Naive Bayes one learned on a dataset of 69 attributes assembled after an attribute selection procedure. Full article

Deep learning techniques play a crucial role in medical image segmentation for diagnostic purposes, with traditional convolutional neural networks (CNNs) and emerging transformers having achieved satisfactory results. CNN-based methods focus on extracting the local features of an image, which are beneficial for handling image details and textural features. However, the receptive fields of CNNs are relatively small, resulting in poor performance when processing images with long-range dependencies. Conversely, transformer-based methods are effective in handling global information; however, they suffer from significant computational complexity arising from the building of long-range dependencies. Additionally, they lack the ability to perceive image details and adopt channel features. These problems can result in unclear image segmentation and blurred boundaries. Accordingly, in this study, a multiscale recombined channel attention (MRCA) module is proposed, which can simultaneously extract both global and local features and has the capability of exploring channel features during feature fusion. Specifically, the proposed MRCA first employs multibranch extraction of image features and performs operations such as blocking, shifting, and aggregating the image at different scales. This step enables the model to recognize multiscale information locally and globally. Feature selection is then performed to enhance the predictive capability of the model. Finally, features from different branches are connected and recombined across channels to complete the feature fusion. Benefiting from fully exploring the channel features, an MRCA-based U-Net (MRCA-UNet) framework is proposed for medical image segmentation. Experiments conducted on the Synapse multi-organ segmentation (Synapse) dataset and the International Skin Imaging Collaboration (ISIC-2018) dataset demonstrate the competitive segmentation performance of the proposed MRCA-UNet, achieving an average Dice Similarity Coefficient (DSC) of 81.61% and a Hausdorff Distance (HD) of 23.36 on Synapse and an Accuracy of 95.94% on ISIC-2018. Full article

In subsurface geological mapping, it is more than advisable to compare different solutions obtained with neural and other algorithms. Here, for such comparison, we used the previously published and well-prepared dataset of subsurface data collected from the Bjelovar Subdepression, a 2900 km² large regional macrounit in the Croatian part of the Pannonian Basin System. Data on depth were obtained for the youngest (the shallowest) Lonja Formation (Pliocene, Quaternary) and mapped using neural network (NN), inverse distance weighting (IDW), and ordinary kriging (OK) algorithms. The obtained maps were compared based on square error (using k-fold cross-validation) and the visual interpretation of isopaches. Two other algorithms were also tested, namely, random forest (RF) and extreme gradient boosting (XGB) algorithms, but they were rejected as inappropriate for this purpose solely based on the visuals of the obtained maps, which did not follow any interpretable geological structures. The results showed that NN is a highly adjustable method for interpolation, with adjustment for numerous hyperparameters. IDW showed its strength as one of the classical interpolators, and its results are always located close to the top if several methods are compared. OK is the relative winner, showing the flexibility of variogram analysis regarding the number of data points and possible clustering. The presented variogram model, even with a relatively high sill and occasional nugget effect, can be well fitted into OK, giving better results than other methods when applied to the presented area and datasets. This was not surprising because kriging is a well-established method used exclusively for interpolation. In contrast, NN and machine learning algorithms are used in many fields, and these algorithms, particularly the fitting of hyperparameters in NN, simply cannot be the best solution for all. Full article

The purpose of this paper is twofold. On a technical side, we propose an extension of the Hausdorff distance from metric spaces to spaces equipped with asymmetric distance measures. Specifically, we focus on extending it to the family of Bregman divergences, which includes the popular Kullback–Leibler divergence (also known as relative entropy). The resulting dissimilarity measure is called a Bregman–Hausdorff divergence and compares two collections of vectors—without assuming any pairing or alignment between their elements. We propose new algorithms for computing Bregman–Hausdorff divergences based on a recently developed Kd-tree data structure for nearest neighbor search with respect to Bregman divergences. The algorithms are surprisingly efficient even for large inputs with hundreds of dimensions. As a benchmark, we use the new divergence to compare two collections of probabilistic predictions produced by different machine learning models trained using the relative entropy loss. In addition to the introduction of this technical concept, we provide a survey. It outlines the basics of Bregman geometry, and motivated the Kullback–Leibler divergence using concepts from information theory. We also describe computational geometric algorithms that have been extended to this geometry, focusing on algorithms relevant for machine learning. Full article

Purpose: This study aims to create a deep learning (DL) model capable of accurately delineating the ventricles, and by extension, the periventricular space (PVS), following the 2021 EPTN Neuro-Oncology Atlas guidelines on T1-weighted contrast-enhanced MRI scans (T1CE). The performance of this DL model was quantitatively and qualitatively compared with an off-the-shelf model. Materials and Methods: An nnU-Net was trained for ventricle segmentation using both CT and T1CE MRI images from 78 patients. Its performance was compared to that of a publicly available pretrained segmentation model, SynthSeg. The evaluation was conducted on both internal (N = 18) and external (n = 18) test sets, with each consisting of paired CT and T1CE MRI images and expert-delineated ground truths (GTs). Segmentation accuracy was assessed using the volumetric Dice Similarity Coefficient (DSC), 95th percentile Hausdorff distance (HD95), surface DSC, and added path length (APL). Additionally, a local evaluation of ventricle segmentations quantified differences between manual and automatic segmentations across both test sets. All segmentations were scored by radiotherapy technicians for clinical acceptability using a 4-point Likert scale. Results: The nnU-Net significantly outperformed the SynthSeg model on the internal test dataset in terms of median [range] DSC, 0.93 [0.86–0.95] vs. 0.85 [0.67–0.91], HD95, 0.9 [0.7–2.5] mm vs. 2.2 [1.7–4.8] mm, surface DSC, 0.97 [0.90–0.98] vs. 0.84 [0.70–0.89], and APL, 876 [407–1298] mm vs. 2809 [2311–3622] mm, all with p < 0.001. No significant differences in these metrics were found in the external test set. However clinical ratings favored nnU-Net segmentations on the internal and external test sets. In addition, the nnU-Net had higher clinical ratings than the GT delineation on the internal and external test set. Conclusions: The nnU-Net model outperformed the SynthSeg model on the internal dataset in both segmentation metrics and clinician ratings. While segmentation metrics showed no significant differences between the models on the external set, clinician ratings favored nnU-Net, suggesting enhanced clinical acceptability. This suggests that nnU-Net could contribute to more time-efficient and streamlined radiotherapy planning workflows. Full article

The permanently shadowed regions of the lunar South Pole have become a key target for international lunar exploration due to their unique scientific value and engineering challenges. In order to effectively screen suitable landing zones near the lunar South Pole, this research proposes a comprehensive evaluation method based on a self-organizing map (SOM). Using multi-source remote sensing data, the method classifies and analyzes candidate landing zones by combining scientific purposes (such as hydrogen abundance, iron oxide abundance, gravity anomalies, water ice distance analysis, and geological features) and engineering constraints (such as Sun visibility, Earth visibility, slope, and roughness). Through automatic clustering, the SOM model finds the important regions. Subsequently, it integrates with a supervised learning model, a random forest, to determine the feature importance weights in more detail. The results from the research indicate the following: the areas suitable for landing account for 9.05%, 5.95%, and 5.08% in the engineering, scientific, and synthesized perspectives, respectively. In the weighting analysis of the comprehensive data, the weights of Earth visibility, hydrogen abundance, kilometer-scale roughness, and slope data all account for more than 10%, and these are thought to be the four most important factors in the automated site selection process. Furthermore, the kilometer-scale roughness data are more important in the comprehensive weighting, which is in line with the finding that the kilometer-scale roughness data represent both surface roughness from an engineering perspective and bedrock geology from a scientific one. In this study, a local examination of typical impact craters is performed, and it is confirmed that all 10 possible landing sites suggested by earlier authors are within the appropriate landing range. The findings demonstrate that the SOM-model-based analysis approach can successfully assess lunar South Pole landing areas while taking multiple constraints into account, uncovering spatial distribution features of the region, and offering a rationale for choosing desired landing locations. Full article

Recent European policies fostering sustainable mobility target urban proximity as a core strategy for a modal shift towards low-carbon modes. Urban proximity, as a characteristic of the built environment, can be studied as a sub-thread of a broad and complex body of literature which associates urban factors such as density or land use mix with observed travel behavior, so as to address their relative influence on the latter. Building on this previous knowledge, the present work addresses the importance of a diverse set of factors on local travel modal choice between walking and other modes, according to the 2018 Household Mobility Survey of the Metropolitan Region of Madrid, and a large variety of demographic and built environment characteristics. The work proposes to address this importance through a workflow on a set of Machine Learning models, filtering different distance thresholds and purposes of the trips, going through a strict feature selection process, and executing under different schema definitions. The resulting models are inspected for accuracy, feature importance, and composition. Results suggest that even small changes in distance thresholds exert a great impact on all models; sociodemographic variables are slightly more important in most models, yet building age, along with other street layout factors, pervasively obtain fairly accurate predictions too. Full article

The development and use of conversational agents in education has become widespread in recent years because of their ability to facilitate student interaction with the learning material, improve engagement and provide academic support, while at the same time reducing the teachers’ workload. This is especially important in the case of distance and asynchronous education, where the availability of academic support must be ensured at any time. This paper reports on the implementation and evaluation of a conversational agent called AIMT Agent, developed for the purposes of supporting postgraduate students during their studies. The conversational agent is based on an open-source framework, namely Rasa, which provides the tools for building natural language understanding into the agent. The agent is fully integrated with the Moodle Learning Management System (LMS). The agent was assessed through a questionnaire according to the Technology Acceptance Model (TAM) in terms of perceived usefulness and perceived ease of use by 24 postgraduate students. The results show that users of the AIMT agent have assessed the conversational agent favorably in both of these aspects. This confirms the validity of the approach and is a motivation for refinements and further development. Full article