Search Results (2,085)

The aim of this study is to develop the next-generation fully autonomous electric wheelchair capable of operating in diverse environments. This study proposes a self-controlled autonomous mobility system that integrates a monocular camera and laser-based 3D spatial recognition, convolutional neural network-based obstacle recognition, shape measurement, and stair structure recognition technology. Obstacle recognition and shape measurement are performed by analyzing the surrounding space using convolutional neural networks and distance calculation methods based on laser measurements. The stair structure recognition technology utilizes the stair-step characteristics from the laser’s irradiation pattern, enabling detection of distance information not captured by the camera. A principal analysis and algorithm development were conducted using a small-scale autonomous mobility system, and its feasibility was determined by application to an omnidirectional self-controlled autonomous electric wheelchair. Using the autonomous robot, we successfully demonstrated an obstacle-avoidance program based on obstacle recognition and shape measurement that is independent of environmental illumination. Additionally, 3D analysis of the number of stair steps, height, and depth was achieved. This study enhances mobility in complex environments under varying lighting conditions and lays the groundwork for inclusive mobility solutions in a barrier-free society. When the proposed method was applied to an omnidirectional self-controlled electric wheelchair, it accurately detected the distance to obstacles, their shapes, as well as the height and depth of stairs, with a maximum error of 0.8 cm. Full article

Learning and understanding of art are increasingly understood as dynamic processes in which emotion and cognition unfold over time. However, classroom-based evidence on how structured temporal intervals and guided prompts reshape students’ emotional experience remains limited. This study addresses these gaps by quantitatively examining changes in emotion over time in a higher education institution. Employing a comparative experimental design, third-year undergraduate art students participated in two structured courses, where emotional responses were captured using an emotion recognition approach (facial expression and self-reported text) during two sessions: initial impression and delayed impression (three days later). The findings reveal a high consistency in dominant facial expressions and substantial agreement in self-reported emotions across both settings. However, the delayed impression elicited greater emotional diversity and intensity, reflecting deeper cognitive engagement and emotional processing over time. These results reveal a longitudinal trajectory of emotion influenced by guided reflective re-view over time. Emotional dynamics extend medium theory by embedding temporal and affective dimensions into TBMA course settings. This study proposes an ethically grounded and technically feasible framework for emotion recognition that supports reflective learning rather than mere measurement. Together, these contributions redefine TBMA education as a temporal and emotional ecosystem and provide an empirical foundation for future research on how emotion fosters understanding, interest, and appreciation in higher media art education. Full article

Star scientists are highly influential researchers who have made significant contributions to their field, gained widespread recognition, and often attracted substantial research funding. They are critical for the advancement of science and innovation and significantly influence the transfer of knowledge and technology to industry. Identifying potential star scientists before their performance becomes outstanding is important for recruitment, collaboration, networking, and research funding decisions. This study utilizes machine learning techniques and builds four different classifiers, i.e., random forest, support vector machines, naïve bayes, and logistic regression, to predict star scientists in the field of artificial intelligence while highlighting features related to their success. The analysis is based on publication data collected from Scopus from 2000 to 2019, incorporating a diverse set of features such as gender, ethnic diversity, and collaboration network structural properties. The random forest model achieved the best performance with an AUC of 0.75. Our results confirm that star scientists follow different patterns compared to their non-star counterparts in almost all the early-career features. We found that certain features, such as gender and ethnic diversity, play important roles in scientific collaboration and can significantly impact an author’s career development and success. The most important features in predicting star scientists in the field of artificial intelligence were the number of articles, betweenness centrality, research impact indicators, and weighted degree centrality. Our approach offers valuable insights for researchers, practitioners, and funding agencies interested in identifying and supporting talented researchers. Full article

Detecting citrus Huanglongbing (HLB) in orchard environments is particularly challenging due to multi-scale targets and occlusions due to clustering, which manifest as complex and variable backgrounds, targets ranging from distant single leaves to nearby full canopies, and frequent instances where symptomatic leaves are hidden behind others, all significantly hindering accurate detection. To overcome these challenges, this study introduces a novel citrus object detection model, Multi-Scale Huanglongbing DETR (MSHLB-DETR), developed on the basis of an improved Real-Time DEtection TRansformer (RT-DETR). The model significantly enhances detection accuracy and efficiency for HLB under complex orchard conditions. To address the issue of small target feature loss in leaf detection, a new efficient transformer module called Smart Disease Recognition for Citrus Huanglongbing with Multi-scale (SDRM) is introduced. SDRM includes a space-to-depth (SPD) module and inverted residual mobile block (IRMB), which facilitate deep interaction between local and global features and significantly improve the computational efficiency of the transformer. Additionally, the transformer encoder incorporates a Context-Guided Block (CGBlock) for contextual feature learning. To evaluate the proposed model under complex background conditions, a dataset of 4367 images was collected from diverse orchard scenes, preprocessed, and divided into training, validation, and testing subsets. The experimental results demonstrate that the proposed MSHLB-DETR achieved the best detection performance on the test set, with an ${\mathrm{mAP}}_{50}$ of 96.0%, surpassing other state-of-the-art models of similar scale. Compared to the original RT-DETR, the proposed model increased ${\mathrm{mAP}}_{50}$ by 15.8%, reduced Params by 7.5%, and decreased GFLOPs by 5.2%. This study reveals the critical importance of developing efficient multi-scale detection techniques for the accurate identification of citrus Huanglongbing in complex real-time monitoring scenarios. The proposed algorithm is expected to provide valuable references and new insights for the precise and timely detection of citrus Huanglongbing. Full article

Weakly Supervised Video Anomaly Detection (WSVAD) is a critical task in computer vision. It aims to localize and recognize abnormal behaviors using only video-level labels. Without frame-level annotations, it becomes significantly challenging to model temporal dependencies. Given the diversity of abnormal events, it is also difficult to model semantic representations. Recently, the cross-modal pre-trained model Contrastive Language-Image Pretraining (CLIP) has shown a strong ability to align visual and textual information. This provides new opportunities for video anomaly detection. Inspired by CLIP, WSVAD-CLIP is proposed as a framework that uses its cross-modal knowledge to bridge the semantic gap between text and vision. First, the Axial-Graph (AG) Module is introduced. It combines an Axial Transformer and Lite Graph Attention Networks (LiteGAT) to capture global temporal structures and local abnormal correlations. Second, a Text Prompt mechanism is designed. It fuses a learnable prompt with a knowledge-enhanced prompt to improve the semantic expressiveness of category embeddings. Third, the Abnormal Visual-Guided Text Prompt (AVGTP) mechanism is proposed to aggregate anomalous visual context for adaptively refining textual representations. Extensive experiments on UCF-Crime and XD-Violence datasets show that WSVAD-CLIP notably outperforms existing methods in coarse-grained anomaly detection. It also achieves superior performance in fine-grained anomaly recognition tasks, validating its effectiveness and generalizability. Full article

The SETD6 protein lysine methyltransferase monomethylates specific lysine residues in a diverse set of substrates which contain the target lysine residue in a highly variable amino acid sequence context. To investigate the mechanism underlying this multispecificity, we analyzed SETD6 substrate recognition using AlphaFold 3 docking and peptide SPOT array methylation experiments. Structural modeling of the SETD6–E2F1 complex suggested that substrate binding alone is insufficient to restrict SETD6 activity to only one lysine residue, pointing to additional sequence readout at the target site. Methylation of mutational scanning peptide SPOT arrays derived from four different SETD6 substrates (E2F1 K117, H2A.Z K7, RELA K310, and H4 K12) revealed sequence preferences of SETD6 at positions −1, +2, and +3 relative to the target lysine. Notably, glycine or large aliphatic residues were favored at −1, isoleucine/valine at +2, and lysine at +3. These preferences, however, were sequence context dependent and variably exploited among different substrates, indicating conformational variability of the enzyme–substrate interface. Mutation of SETD6 residue L260, which forms a contact with the +2 site in the available SETD6-RELA structure, further demonstrated substrate-specific differences in recognition at the +2/+3 sites. Together, these findings reveal a versatile mode of peptide recognition in which the readout of each substrate position depends on the overall substrate peptide sequence. These findings can explain the multispecificity of SETD6 and similar mechanisms may underlie substrate selection in other protein methyltransferases. Full article

Sepsis remains a critical medical emergency caused by a dysregulated immune response to infection, with timely detection and intervention being essential for improving survival rates. Traditional methods often rely on clinician intuition and structured scoring systems, which may be time-intensive and prone to variability. To address these limitations, Machine Learning (ML) offers a powerful alternative, bringing precision and efficiency to sepsis detection. This study investigates both white-box and complex black-box ML models applied to patient data collected across the continuum of care, including monitoring at the urgent care, en route in ambulances, and diagnostics conducted within hospital emergency department settings themselves. White-box models, such as logistic regression and decision trees, are valued for their interpretability, allowing healthcare providers to understand and trust the reasoning behind predictions. Meanwhile, black-box models like deep neural networks and support vector machines deliver superior accuracy but pose challenges in clinical transparency. This trade-off between explainability and performance is explored in detail, supported by experimental results aimed at identifying the most effective computational strategies for early sepsis recognition across diverse healthcare environments. Full article

Background/Objectives: Forensic dental determination plays a central role in human identification, age estimation, and trauma analysis in medico-legal contexts. Traditional approaches—including clinical examination, odontometric analysis, and radiographic comparison—remain essential but are constrained by examiner subjectivity, population variability, and reduced applicability in fragmented or degraded remains. Recent advances in cone-beam computed tomography (CBCT), three-dimensional surface scanning, intraoral imaging, and artificial intelligence (AI) offer promising opportunities to enhance accuracy, reproducibility, and integration with multidisciplinary forensic evidence. The aim of this review is to synthesize conventional and emerging approaches in forensic odontology, critically evaluate their strengths and limitations, and highlight areas requiring validation. Methods: A structured literature search was performed in PubMed, Scopus, Web of Science, and Google Scholar for studies published between 2015 and 2025. Search terms combined forensic odontology, dental identification, CBCT, 3D scanning, intraoral imaging, and AI methodologies. From 108 records identified, 81 peer-reviewed articles met eligibility criteria and were included for analysis. Results: Digital methods such as CBCT, 3D scanning, and intraoral imaging demonstrated improved diagnostic consistency compared with conventional techniques. AI-driven tools—including automated age and sex estimation, bite mark analysis, and restorative pattern recognition—showed potential to enhance objectivity and efficiency, particularly in disaster victim identification. Persistent challenges include methodological heterogeneity, limited dataset diversity, ethical concerns, and issues of legal admissibility. Conclusions: Digital and AI-based approaches should complement, not replace, the expertise of forensic odontologists. Standardization, validation across diverse populations, ethical safeguards, and supportive legal frameworks are necessary to ensure global reliability and medico-legal applicability. Full article

Background: Most immunologically “cold” tumors do not respond durably to checkpoint blockade because tumor antigen (TA) release and presentation are insufficient to prime effective T-cell immunity. While prior work demonstrated synergy between cisplatin and a TLR7/8/9 agonist (CR108) in 4T1 tumors, the underlying mechanism—particularly whether chemotherapy functions as a broad antigen-releasing agent enabling TLR-driven immune amplification—remained undefined. Methods: Using murine models of breast (4T1), melanoma (B16-F10), and colorectal cancer (CT26), we tested multiple chemotherapeutic classes combined with CR108. We quantified intratumoral and systemic soluble TAs, antigen presentation and cross-priming by antigen-presenting cells, tumor-infiltrating lymphocytes, and cytokine production by flow cytometry/ICS. T-cell receptor β (TCRβ) repertoire dynamics in tumor-draining lymph nodes were profiled to assess amplitude and breadth. Tumor microenvironment remodeling was analyzed, and public datasets (e.g., TCGA basal-like breast cancer) were interrogated for expression of genes linked to TA generation/processing and peptide loading. Results: Using cisplatin + CR108 in 4T1 as a benchmark, we demonstrate that diverse chemotherapies—especially platinum agents—broadly increase the repertoire of soluble tumor antigens available for immune recognition. Across regimens, chemotherapy combined with CR108 increased T-cell recognition of candidate TAs and enhanced IFN-γ⁺ CD8⁺ responses, with platinum agents producing the largest expansions in soluble TAs. TCRβ sequencing revealed increased clonal amplitude without loss of repertoire breadth, indicating focused yet diverse antitumor T-cell expansion. Notably, therapeutic efficacy was not predicted by canonical damage-associated molecular pattern (DAMP) signatures but instead correlated with antigen availability and processing capacity. In human basal-like breast cancer, higher expression of genes involved in TA generation and antigen processing/presentation correlated with improved survival. Conclusions: Our findings establish an antigen-centric mechanism underlying chemo–TLR agonist synergy: chemotherapy liberates a broadened repertoire of tumor antigens, which CR108 then leverages via innate immune activation to drive potent, T-cell-mediated antitumor immunity. This framework for rational selection of chemotherapy partners for TLR7/8/9 agonism and support clinical evaluation to convert “cold” tumors into immunologically responsive disease. Full article

Indonesia boasts diverse tropical fruits like ciplukan, harendong, and kecapi, which are nutrient-rich but underutilized. To address this, an automated fruit recognition system was developed using Convolutional Neural Network (CNN) with MobileNet architecture, leveraging Depthwise Separable Convolution (DSC) for efficiency. The model was trained on 5000 images of five local fruits (224 × 224 pixels), split into 70% training, 20% validation, and 10% testing. Optimized for Android, the system enables real-time identification with minimal hardware requirements (e.g., 2 GB RAM for low-end devices). Evaluation metrics (accuracy, precision, recall) achieved 97.43% accuracy, demonstrating MobileNet’s effectiveness. This study highlights deep learning’s potential in preserving and promoting Indonesia’s indigenous fruits. Full article

This paper proposes TWISS (TOPSIS + Wrapper Incremental Subset Selection), a novel hybrid feature selection framework designed for electromyographic (EMG) pattern recognition in upper-limb prosthetic control. TWISS integrates the multi-criteria decision-making method TOPSIS with a forward wrapper search strategy, enabling subject-specific feature optimization based on a ranking that combines filter metrics, including Chi-squared, ANOVA, and Mutual Information. Unlike conventional static feature sets, such as the Hudgins configuration (48 features: four per channel, 12 channels) or All Features (192 features: 16 per channel, 12 channels), TWISS dynamically adapts feature subsets to each subject, addressing inter-subject variability and classification robustness challenges in EMG systems. The proposed algorithm was evaluated on the publicly available Ninapro DB7 dataset, comprising both intact and transradial amputee participants, and implemented in an open-source, fully reproducible environment. Two Google Colab tools were developed to support diverse workflows: one for end-to-end feature extraction and selection, and another for selection on precomputed feature sets. Experimental results demonstrated that TWISS achieved a median F1-macro score of 0.6614 with Logistic Regression, outperforming the All Features set (0.6536) and significantly surpassing the Hudgins set (0.5626) while reducing feature dimensionality. TWISS offers a scalable and computationally efficient solution for feature selection in biomedical signal processing and beyond, promoting the development of personalized, low-cost prosthetic control systems and other resource-constrained applications. Full article

People with Autism Spectrum Disorder (ASD) face significant barriers to accessing and maintaining employment, many of which stem from work environments that fail to accommodate their neurological diversity. This article aims to analyze the occupational risks faced by autistic individuals in the workplace. A total of 39 scientific studies were reviewed, and the results identified nine predominant thematic categories of occupational risks. Particularly prominent were deficient communication, lack of structured support, cognitive overload, and difficulties coping with change. The reported situations were examined in detail, with attention paid to their specific contexts. A clear predominance of psychosocial risks over ergonomic ones was observed. The review also highlights several underexplored yet equally relevant risk factors, such as discontinuity in supported employment programs, difficulties in requesting reasonable accommodations, discrimination, a lack of professional recognition, and the negative effects of digital or remote environments, such as isolation. This study underscores the importance of recognizing unsafe conditions arising from the lack of neurodiversity-informed adjustments as a necessary step toward implementing organizational and social adaptations in the workplace. Full article

To address the challenges of poor model generalization and suboptimal recognition accuracy stemming from limited and imbalanced sample sizes in tomato leaf disease identification, this study proposes a novel recognition strategy. This approach synergistically combines an enhanced image augmentation method based on generative adversarial networks with a lightweight deep learning model. Initially, an Atrous Spatial Pyramid Pooling (ASPP) module is integrated into the CycleGAN framework. This integration enhances the generator’s capacity to model multi-scale pathological lesion features, thereby significantly improving the diversity and realism of synthesized images. Subsequently, the Convolutional Block Attention Module (CBAM), incorporating both channel and spatial attention mechanisms, is embedded into the MobileNetV4 architecture. This enhancement boosts the model’s ability to focus on critical disease regions. Experimental results demonstrate that the proposed ASPP-CycleGAN significantly outperforms the original CycleGAN across multiple disease image generation tasks. Furthermore, the developed CBAM-MobileNetV4 model achieves a remarkable average recognition accuracy exceeding 97% for common tomato diseases, including early blight, late blight, and mosaic disease, representing a 1.86% improvement over the baseline MobileNetV4. The findings indicate that the proposed method offers exceptional data augmentation capabilities and classification performance under small-sample learning conditions, providing an effective technical foundation for the intelligent identification and control of tomato leaf diseases. Full article

Cleptoparasitic bees of the subgenus Stelis (Stelidomorpha) occur mainly in the Mediterranean and Middle East. In this study, we elevate Stelis aegyptiaca ssp. canaria Warncke, 1992 to species rank (S. canaria Warncke, 1992) and describe two new species, Stelis alainensis Kasparek sp. nov. and Stelis surica Kasparek sp. nov., both discovered in Oman and the United Arab Emirates. Morphological differences between these species and their closest relatives (S. aegyptiaca Radoszkowski, 1876, S. pentelica Mavromoustakis, 1963, and S. nasuta (Latreille, 1809)) are corroborated by genetic divergence in the mitochondrial COI barcode region, with Kimura 2-parameter (K2P) distances of 7.6–15.2%. A notable case is Stelis nasuta, which shows deep genetic subdivision into three clusters: (1) Iberian Peninsula and North Africa, (2) southeastern France, Italy, and the Balkans, (3) eastern Balkans, Turkey, and the Levant. Moderate genetic K2P distances of 2.9–3.3% complicated species delimitation. Analyses with ABGD, ASAP, bPTP, and RESL algorithms consistently supported recognition of these lineages as putative species. As multivariate analyses of 11 morphometric traits revealed no consistent diagnostic differences, we treat these lineages as phylospecies rather than formal taxa. Our findings demonstrate that bee diversity in the Palaearctic remains underestimated, and that expanded sampling and integrative approaches continue to reveal hidden lineages. Full article

This paper focuses on the efficient fine-tuning of large language models within the federated learning framework. To address the performance bottlenecks caused by multi-source heterogeneity and structural inconsistency, a structure-aware federated fine-tuning method is proposed. The method incorporates a graph representation module (GRM) to model internal structural relationships within text and employs a segmentation mechanism (SM) to reconstruct and align semantic structures across inputs, thereby enhancing structural robustness and generalization under non-IID (non-Independent and Identically Distributed) settings. During training, the method ensures data locality and integrates structural pruning with gradient encryption (SPGE) strategies to balance privacy preservation and communication efficiency. Compared with representative federated fine-tuning baselines such as FedNLP and FedPrompt, the proposed method achieves consistent accuracy and F1-score improvements across multiple tasks. To evaluate the effectiveness of the proposed method, extensive comparative experiments are conducted across tasks of text classification, named entity recognition, and question answering, using multiple datasets with diverse structures and heterogeneity levels. Experimental results show that the proposed approach significantly outperforms existing federated fine-tuning strategies on most tasks, achieving higher performance while preserving privacy, and demonstrating strong practical applicability and generalization potential. Full article