Search Results (6,206)

Small object detection in UAV remote sensing imagery plays a crucial role in applications such as infrastructure inspection, disaster assessment, and precision agriculture, where targets of interest frequently occupy fewer than $32\times 32$ pixels under large ground sampling distance variation and complex cluttered backgrounds. Existing methods still face three main challenges in UAV small-object detection: fine-grained detail loss caused by repeated downsampling, feature inconsistency during cross-scale fusion, and unstable boundary regression in densely distributed aerial scenes. To address these issues, this paper proposes HSAR-DETR, a detection framework that jointly improves hierarchical feature representation, cross-scale refinement, and geometry-aware localization. Specifically, a Hierarchical Enhancement Network (HENet) is introduced to preserve shallow spatial details while strengthening deep semantic-context representation. A Dual-Stream Feature Refinement module (DSFR) is designed at the P4-to-P3 fusion stage, combining spatial-domain structural modeling with frequency-domain phase refinement to improve cross-scale feature consistency. A Coordinate-Guided Adaptive Convolution module (CGAC) is further deployed before the detection head, converting coordinate-guided offset magnitudes into modulation weights for adaptive feature recalibration and improved localization stability. In addition, a conventional high-resolution P2 detection branch is incorporated to enhance small-object representation. Experimental results on the VisDrone, RSOD, and TinyPerson datasets demonstrate improved detection performance. On the VisDrone validation set, HSAR-DETR achieves 50.8% $mA{P}_{50}$ and 31.4% $mA{P}_{50:95}$, outperforming the RT-DETR baseline by 4.2 and 3.0 percentage points, respectively. Full article

The rapid growth of organisational textual data has increased the value of Natural Language Processing (NLP) and semantic analytics for strategic decision support. However, many employees still lack the knowledge and skills needed to evaluate AI-generated information critically. This study develops an explainable Management Information Systems (MIS) framework that integrates NLP-based semantic analytics with PLS-SEM to examine the relationship between AI illiteracy and strategic decision quality. A convergent mixed-methods design with sequential analytical integration was used with a valid sample of 200 knowledge workers from public organisations in Saudi Arabia across six industries. The sample included employees from Saudi Arabia, Egypt, Jordan, Sudan, Syria, India, and the Philippines. Quantitative data were analysed using PLS-SEM, while textual data were analysed using Sentence-BERT, BERTopic, semantic network analysis, and Aspect-Based Sentiment Analysis. The results showed that higher AI illiteracy was negatively associated with strategic decision quality and positively associated with automation bias, uncritical trust in AI, and cognitive offloading. Digital proficiency and AI governance awareness weakened the negative association between AI illiteracy and decision quality, while functional-background differences were examined through multigroup analysis. The semantic analysis identified six themes: AI competency, decision trust, AI governance, decision support, organisational learning, and risk awareness. Sentiment analysis showed positive views of productivity and decision support, together with concerns about algorithmic bias, explainability, transparency, and AI governance. The study contributes an integrated human–AI decision vulnerability framework in which semantic evidence complements structural modelling and provides a clearer understanding of AI-related competency, reliance, governance, and decision-support issues. Full article

Accurate brain tumor segmentation from multi-modal magnetic resonance imaging (MRI) is essential for clinical diagnosis and treatment planning. However, effectively capturing long-range contextual information and fine lesion boundaries under limited computational budgets remains challenging. In this work, we propose SDFR-Net, a lightweight stage-asymmetric Spectral Diffusion and Frequency–Spatial Refinement Network for efficient 2.5D brain tumor MRI segmentation. Instead of applying identical processing across all hierarchical stages, SDFR-Net adopts stage-dependent spectral diffusion, stage-selective conditional refinement, and asymmetric cross-stage frequency-grid allocation to accommodate the distinct semantic and frequency characteristics of shallow and deep representations. The network consists of a Spectral Diffusion Encoder for spectral-domain contextual propagation, a Frequency–Spatial Enhancement Module for adaptive refinement of multi-scale skip features, and a lightweight Conditional Refinement Decoder for lesion-aware reconstruction. Experiments on the BraTS 2019 and BraTS 2020 datasets demonstrate that SDFR-Net achieves whole-tumor Dice scores of 0.856 and 0.880, respectively, while requiring only 1.33 M parameters. Ablation comparisons of stage-selective FiLM injection and symmetric versus asymmetric frequency-grid schedules further support the stage-asymmetric design. These results indicate that SDFR-Net provides a favorable accuracy–efficiency trade-off for resource-constrained brain tumor MRI segmentation. Full article

Financial media risk is jointly driven by multisource content, including news reports, corporate announcements, social media posts, short videos, and livestreams, while content authenticity, propagation velocity, asset relevance, and trading infrastructure conditions can simultaneously influence short-term market fluctuations. To address the limitations of existing methods, including their reliance on either textual information or market sequences, insufficient source verification, and inadequate alignment of asynchronous multimodal signals, FMRP-Net is proposed for artificial intelligence-driven sensing. Event semantics, risk categories, and asset association information are first extracted through a pretrained financial language model and cross-modal consistency analysis. A dual-layer hardware reliability perception module is then employed to integrate sensing evidence from cameras, microphones, terminal inertial signals, server temperature, power consumption, network traffic, and transmission latency. Heterogeneous temporal propagation graphs, cross-frequency alignment, and bidirectional propagation–market coupling are further incorporated to jointly predict market direction, volatility, risk level, and propagation trends. Experimental results demonstrate that FMRP-Net achieved an Accuracy of $0.832$, a Macro-F1 of $0.824$, a ROC-AUC of $0.891$, an MCC of $0.665$, and a PR-AUC of $0.883$ for market direction prediction over future horizons of 5, 15, 30, and 60 min, indicating a balanced performance in terms of Precision and Recall. For volatility prediction, MAE, RMSE, and MAPE values of $0.0178$, $0.0271$, and $12.46\%$ were obtained, respectively, together with an $R^2$ of $0.812$. In the ablation study, the media risk Macro-F1 and source reliability AUC reached $0.842$ and $0.929$, respectively, while the propagation-scale prediction error was reduced to $0.109$ and the average early-warning lead time reached $10.6$ min. These results demonstrate that the integration of multimedia semantics, hardware sensing evidence, and propagation structures can effectively improve the accuracy, stability, and interpretability of financial market prediction and risk early warning. Full article

Short-term financial risk in intelligent trading systems is reflected not only in prices, trading volumes, and textual sentiment but also in infrastructure operating states, including server workload, device power consumption, network latency, and packet loss rate. We propose HSF-LLMNet, a hardware sensor and large language model semantic fusion network for jointly modeling external information shocks and infrastructure responses. A large language model extracts event category, sentiment polarity, risk intensity, and semantic uncertainty from financial texts. Reliability-aware temporal modeling handles sensor missingness, drift, and abnormal noise, while asynchronous soft alignment, bidirectional cross-attention, and reliability-aware gated fusion integrate irregular textual events with continuous hardware signals. The model jointly predicts market direction, realized volatility, and three-level risk over the subsequent 30 min. Experiments were conducted on eight Chinese A-share indices: the SSE Composite Index (000001.SH), SSE 50 Index (000016.SH), CSI 300 Index (000300.SH), STAR 50 Index (000688.SH), CSI 500 Index (000905.SH), CSI 1000 Index (000852.SH), Shenzhen Component Index (399001.SZ), and ChiNext Index (399006.SZ). The common observation period for market, textual, and hardware data extended from 1 March 2024 to 30 June 2025. After data cleaning, timestamp matching, and multimodal temporal alignment, 169,208 aligned asset–time prediction windows were retained for the 30 min forecasting task. Realized volatility was defined as the square root of the sum of squared one-minute log returns over the future 30 min interval. The three-level risk label was constructed from future realized volatility, absolute 30 min return, and liquidity stress, with all thresholds estimated exclusively from the training portion of each fold. A sample was labeled high risk when at least two of the three indicators exceeded their 85th-percentile thresholds or when any indicator exceeded its 95th-percentile threshold. It was labeled medium risk when, after excluding high-risk samples, at least two indicators exceeded their 60th-percentile thresholds or any indicator exceeded its 85th-percentile threshold; all remaining samples were labeled low risk. Results showed that HSF-LLMNet achieved an accuracy of 78.62%, a precision of 78.14%, a recall of 77.83%, an F1-score of 77.98%, an area under the receiver operating characteristic curve of 84.91%, and a Matthews correlation coefficient of 57.36% for directional prediction. For realized-volatility regression, the MAE, RMSE, MAPE, and $R^2$ were 0.0089, 0.0135, 9.21%, and 0.812, respectively. For high-risk-event warning, the mean effective warning time, defined as the interval between the first valid alarm and the corresponding event, was 15.37 min; the false-alarm rate and missed-alarm rate were 6.82% and 8.14%, respectively. Ablation experiments showed performance reductions after removing semantic encoding, sensor-reliability estimation, asynchronous alignment, bidirectional cross-attention, gated fusion, or multi-task learning. These results indicate that textual events and infrastructure operating states provide complementary information for quantitative risk analytics and fintech applications. Full article

The rapid development of large model-driven agent applications, such as digital assistants and robots, requires 6G radio access networks (RAN) to deliver enhanced flexibility, adaptability, and low-latency capabilities. However, the existing RAN user plane (UP) architecture suffers from coarse decoupling granularity and significant cross-layer functional redundancy. These limitations severely hinder the on-demand orchestration and dynamic reconfiguration required by heterogeneous agent services. To address these challenges, this paper proposes a ComBERT-driven service-based RAN UP decoupling method, specifically targeting the functional coupling and redundancy between the PDCP and RLC sublayers. First, we develop a domain-specific language model, ComBERT, by pre-training a BERT model on a 3GPP protocol corpus and fine-tuning it on text-matching tasks to deeply comprehend protocol semantics. Subsequently, ComBERT is utilized to extract semantic features from UP functional components, employing a sliding window mechanism to overcome truncation in lengthy protocol texts and using cosine similarity to measure functional relevance. Finally, a threshold-based fusion algorithm is designed to identify and merge cross-layer redundant functions, thereby forming independent service units with distinct responsibilities. These fused units serve as the basic building blocks for scenario-specific orchestration. Simulation results demonstrate that the proposed method reduces the number of UP components by 12.5%, 18.7%, and 18.2% in eMBB, URLLC, and mMTC scenarios, respectively. Simultaneously, it decreases average processing delays by 7.9%, 10.2%, and 11.0% across these respective scenarios. Ultimately, this approach effectively improves the lightweight deployment, processing efficiency, and reconfiguration capabilities of the service-based UP, providing a crucial foundation for on-demand service orchestration in 6G networks tailored to agent services. Full article

Fine-grained poisonous mushroom recognition in natural scenes is challenging because of complex backgrounds, subtle morphological differences, and the limited interpretability of model decisions. To address these challenges, this paper proposes Att-FiLM, an attention-enhanced multi-scale Feature-Wise Linear Modulation network for poisonous mushroom image recognition. The model adopts an asymmetric dual-backbone architecture in which a frozen ConvNeXt-Base branch provides global semantic priors, while a trainable EfficientNet-B0 branch learns local discriminative features. Rather than directly concatenating heterogeneous features, Att-FiLM generates scale and shift parameters from semantic features and performs channel-wise modulation on multi-scale EfficientNet features at Stage 2 and Stage 4. This mechanism enables global semantic information to guide local feature learning while reducing feature redundancy and semantic inconsistency. Experimental results show that Att-FiLM achieves an Accuracy of 95.58% and an F1-score of 0.9455 on the poisonous/edible binary classification task. On the 190-class species-level classification task, it achieves a Top-1 Accuracy of 93.63% and a Macro-F1 of 0.9347. Interpretability analysis further shows that decision-relevant responses are frequently associated with morphologically relevant regions, including gills, annuli, volvae, and cap textures. These results indicate that Att-FiLM provides effective recognition performance together with interpretable decision evidence for mushroom recognition in complex natural scenes. Full article

Pixel-level annotation of remote sensing imagery is costly, motivating weakly supervised semantic segmentation (WSSS) using only image-level labels. However, class activation maps (CAMs) often highlight only discriminative sub-regions and fail to separate adjacent land-cover regions, particularly in remote sensing scenes characterized by densely co-occurring land-cover classes and substantial variations in object scale. To address these limitations, we propose a three-stage framework that integrates complementary priors from Contrastive Language–Image Pre-training (CLIP), Self-Distillation with No Labels version 2 (DINOv2), and the Segment Anything Model (SAM). First, a lightweight CLIP adapter aligns vision–language priors with remote sensing imagery, while sigmoid-based multi-label decoupled distillation replaces class-competitive distillation with independent class-wise supervision, producing more complete CAMs. Second, DINOv2-guided feature clustering decomposes large merged regions before SAM prompt generation, while Spatial–Semantic Constraints are used to construct confidence-guided point-and-box prompts and reject excessively expanded or semantically inconsistent masks, thereby generating reliable pseudo-labels. Finally, a compact segmentation network is initialized with the weights learned in Stage 1 and retrained using the refined pseudo-labels generated in Stage 2, eliminating the need for foundation models during inference. Experiments on the Potsdam, LoveDA, and DeepGlobe datasets show that the proposed method achieves mean intersection over union (mIoU) scores of 53.16%, 52.66%, and 62.98%, respectively, outperforming state-of-the-art WSSS baselines by 6.55, 1.16, and 1.27 percentage points, respectively. These results demonstrate the effectiveness and generalizability of the proposed framework across diverse remote sensing scenarios under image-level supervision. Full article

Sustainable historic-town conservation requires preserving historical patterns and cultural memory while understanding how spatial environments shape everyday perception. Using Yun’an Ancient Salt Town in Chongqing, China, this study develops a three-layer framework of spatial topology, scene interfaces, and historical semantic elements. Space syntax, deep-learning image recognition, rule-based coding, image-based questionnaires, Spearman correlation, random forest regression, and SHAP were used to examine perceived historicity, safety, attractiveness, and comfort. The analysis combined axial models for 1985, 2004, and 2024 with 140 images—55 human view and 85 aerial view—each evaluated by 264 valid respondents. Yun’an’s street network shifted from a salt-production and transport structure toward modern traffic corridors; integration declined in the historic core, and intelligibility fell from 0.13 in 2004 to 0.07 in 2024. At the human-view scale, modern interference was negatively associated with historicity and attractiveness, whereas stairs, traditional components, and micro-historical objects were positively associated with historicity. At the aerial-view scale, historical visibility, character integrity, and blue-green-space indicators were associated with historicity, attractiveness, and comfort. Cross-validated random forest performance was strongest for aerial-view historicity and limited for other outcomes. The findings support a human-centered, culturally sustainable renewal pathway linking historical-street continuity, townscape integration, blue-green quality, and heritage-node activation. Full article

With the rapid evolution of next-generation mobile communications, semantic communication has emerged as an intelligent communication paradigm capable of surpassing the Shannon limit. A fundamental prerequisite for this paradigm is the synchronization of background knowledge between the transmitter and receiver, making the semantic knowledge base (SKB) a critical cornerstone. However, effectively selecting appropriate content from massive cloud-based knowledge repositories for edge deployment remains a significant challenge. This paper conducts systematic research to address the key issues in the flow deployment of SKBs at the edge, including insufficient adaptation to personalized preferences, inadequate timeliness management, and the complexity of multi-objective optimization. First, a comprehensive system model is constructed, integrating user preferences, knowledge relevance, transceiver matching degree, and the Age of Information (AOI). Second, the Generative Adversarial Network (GAN)-assisted Preference-based Reinforcement Learning (GaPbRL) algorithm is proposed. The experimental results demonstrate that this method outperforms traditional schemes in terms of knowledge-base hit rate, transceiver matching degree, and algorithm convergence speed, while significantly reducing the overhead of manual fine-tuning. This study provides a robust framework for the personalized and efficient cloud–edge collaborative deployment of SKBs. Full article

Educational assessment systems have primarily relied on episodic forms of assessment, including examinations, assignments, grades, and credentials. These approaches provide efficient and scalable summaries of achievement and yet capture only part of the developmental process through which learners build competence. Moreover, learning increasingly unfolds across digital platforms, workplaces, collaborative networks, and AI-mediated environments, generating rich evidence of learner development that remains fragmented across systems and contexts. Advances in artificial intelligence, learning analytics, multimodal analytics, learner modeling, and semantic interoperability make it increasingly feasible to connect, integrate, and interpret this evidence across contexts and over time. This paper introduces the AI-Mediated Continuous Assessment Infrastructure (AIM-CAI), a sociotechnical framework supporting longitudinal, probabilistic interpretation of distributed evidence of learning. Within AIM-CAI, continuous assessment refers to the ongoing accumulation and dynamic interpretation of evidence generated through learning activities. The framework integrates distributed evidence systems, evidence serialization mechanisms, AI-mediated semantic translation, probabilistic learner models, dynamic competency profiles, and federated governance architectures to support context-sensitive interpretations of learner development while maintaining human judgment, privacy, accountability, and learner agency. The authors examine implications for assessment, credentialing, lifelong learning, institutional roles, interoperability, and governance and outline a research agenda addressing key psychometric, ethical, and governance challenges, including validity, fairness, surveillance, semantic instability, and ownership of learning evidence. Full article

Renewable energy and energy storage stations require frequent updates of monitoring and control logic across heterogeneous devices and changing operating strategies. This paper proposes a rule graph-based reference architecture that combines low-code logic configuration, graph–model semantic binding, and microservice-oriented functional decomposition. A component status matrix separates the target architecture from the implemented subset. The runnable subset comprises a minimal FastAPI backend, REST/WebSocket telemetry interfaces, an in-process queue, and stateful rule evaluators; gateway, authentication, external message bus, time-series database, visual editor, and industrial protocol services remain design-level elements. Beyond the original single-rule example, a priority-ordered multi-device rule is implemented for cooperative BESS dispatch, communication/topology blocking, low-SOC protection, frequency-based load shedding, backup request, and five-sample recovery release. Existing local network benchmarks are complemented by a 600-step software-in-the-loop trace with scripted telemetry fluctuations and communication quality faults and by 500 in-process ASGI timing samples at each of the four point levels. The trace produced no safety dispatch or protected device violations. P99 application path latency ranged from 1.1962 to 5.5287 ms, but one 75.3065 ms outlier exceeded a 50 ms reference deadline, demonstrating that the Windows/FastAPI path is not deterministic. No industrial controller, hardware-in-the-loop facility, field data, or engineer usability study was used. Accordingly, the paper makes no claim of industrial real-time readiness or measured development effort reduction. Full article

High-resolution remote sensing semantic segmentation remains challenging because repeated downsampling progressively weakens the fine-grained spatial information of small objects, while direct fusion of heterogeneous multi-scale features may introduce semantic discrepancies and redundant background responses. To address these issues, this study proposes SEMU-Net, a structure-enhanced multi-branch U-shaped network. First, an independent multi-scale complementary branch is constructed outside the main encoder pathway to provide auxiliary hierarchical representations and compensate for information degradation during progressive semantic abstraction. Second, a scale-consistent feature embedding module is introduced to project and normalize side-branch features before residual injection, thereby improving the compatibility of cross-path feature representations. Third, a discriminative channel modulation module is incorporated into the decoder to adaptively strengthen task-relevant channel responses and suppress redundant background activations. Experiments were conducted on the ISPRS Vaihingen dataset and a self-annotated high-resolution remote sensing dataset. On the Vaihingen dataset, SEMU-Net achieved a mIoU of 72.21% and an Average F1 score of 83.64%, outperforming the strongest competing method by 0.59 and 0.44 percentage points, respectively. The IoU of the Car class increased by 3.50 percentage points. On the self-annotated dataset, the IoU of the narrow Road class improved by 4.12 percentage points. These results demonstrate that SEMU-Net improves overall segmentation accuracy and enhances the recognition of small objects, with the observed improvements being consistent with the design objectives of multi-scale information compensation, cross-path feature adaptation, and channel recalibration. Full article

With the development of smart classrooms and digital learning platforms, multimodal learning behavior data provide a new sensing basis for understanding students’ cognitive states and psychological risk warnings. However, existing educational data mining methods mainly focus on performance prediction, dropout warning, or surface-level emotion recognition, while continuous and interpretable modeling of deeper cognitive biases and related psychological risks remains insufficient. To address this issue, we propose MLBS-Net, a multimodal learning behavior sensing network for teaching feedback that jointly models students’ textual expressions, behavioral sequences, classroom interactions, and psychological auxiliary signals. MLBS-Net integrates theory-guided textual cognitive bias encoding, temporal behavioral state modeling, and reliability-aware cross-modal fusion to capture psychologically interpretable cognitive patterns, characterize dynamic learning-state changes, and adaptively integrate multimodal information according to data quality and task contribution while providing interpretable feedback for teachers. Experimental results show that MLBS-Net achieves a Macro-F1 of 0.855 for cognitive bias recognition and an AUC of 0.891 for psychological risk warning, outperforming traditional machine learning, unimodal deep learning, and standard multimodal methods. Ablation results further support the effectiveness of theory-guided semantic encoding, temporal behavioral modeling, reliability estimation, and multitask learning. These findings demonstrate that MLBS-Net can jointly characterize cognitive biases and potential psychological risks from multisource learning behaviors, providing a feasible approach for learning-state sensing, risk warning, and interpretable teaching support in smart education. Full article

The transition toward urban-scale digital twin and smart city management requires survey-grade 3D basemaps, yet conventional documentation remains time-consuming and prone to inaccuracies. While Terrestrial Laser Scanning (TLS) offers rapid 3D acquisition, capturing complex, GNSS-denied multi-story interiors frequently causes cumulative registration errors and isolated indoor–outdoor data silos. To address this, this study proposes a comprehensive typology-agnostic framework for developing high-fidelity digital twin basemaps. Treating the building as a unified spatial network, the methodology systematically mitigates error propagation through strategic linkage planning, rigid shell-first registration, continuous vertical core anchoring (via stairwells), and adaptive multi-source data fusion. Implemented on an adaptively reused heritage building, the developed basemap achieved an absolute georeferencing accuracy of 30.0 mm (RMSE) against an independent total station control network, alongside a mean relative error of 3.11 mm. Comparative analysis against legacy 2D CAD floor plans revealed simplified geometric representations and categorical dimensional deviations of up to 29.8 cm. Demonstrating its practical utility, the point cloud-centric geometric hub avoids forced geometric idealization, successfully supporting direct immersive visualization, architectural slicing (floor plans, sections, elevations), and multi-LOD algorithmic planar segmentation. This spatially constrained acquisition strategy bypasses legacy limitations, delivering a mathematically verified 3D reality capture essential for smart facility management, heritage conservation, and downstream semantic intelligence. Full article