Search Results (1,658)

Unmanned aerial vehicle (UAV) imagery poses persistent challenges for object detection, including dense small objects, large-scale variation, cluttered backgrounds, and stringent localization requirements, where conventional two-stage detectors often fall short in fine-grained small-object representation, efficient global context modeling, and classification–localization consistency. We specifically target low-altitude UAV-captured imagery with highly flexible viewpoints (near-nadir to oblique) and frequent platform-induced motion blur, which makes dense small-object localization substantially more challenging than in conventional remote-sensing imagery. To address these issues, we propose CASA-RCNN, a context-adaptive and scale-aware two-stage detection framework tailored to UAV scenarios. CASA-RCNN introduces a shallow-level enhancement module, ConvSwinMerge, which strengthens position-sensitive cues and suppresses background interference by combining coordinate attention with channel excitation, thereby improving discriminative high-resolution features for small objects. For deeper semantic features, we incorporate an adaptive sequence modeling module based on MambaBlock to capture long-range dependencies and support context reasoning in crowded or occluded scenes with practical computational overheadon a desktop GPU. In addition, we adopt Varifocal Loss for quality-aware classification to better align confidence scores with localization quality, and we design a ScaleAdaptiveLoss to dynamically reweight regression objectives across object scales, compensating for the reduced gradient contribution of small targets during training. Experiments on the VisDrone2021 validation benchmark show that CASA-RCNN achieves 22.9% mAP, improving Faster R-CNN by 9.0 points; it also reaches 36.6% mAP${}_{50}$ and 25.7% mAP${}_{75}$. Notably, performance on small objects improves to 12.5% mAP${}_s$ (from 6.9%), and ablation studies confirm the effectiveness and complementarity of the proposed components. Full article

Maritime target detection under complex adverse weather conditions (e.g., fog, rain, and low light) is crucial for Unmanned Surface Vehicle (USV) navigation. However, achieving high detection accuracy and efficiency remains challenging due to coupled environmental interference and limited computing resources. In this paper, we propose W-MTD, a task-specific distillation framework designed for weather-robust and lightweight maritime target detection based on knowledge distillation. Building upon the Fine-grained Distribution Refinement (D-FINE) detection model, this method constructs a dual-path knowledge distillation framework tailored for maritime scenes. Through the synergistic optimization of feature similarity constraints and decoupled distillation, it facilitates multi-level knowledge transfer from a teacher model to a lightweight student model, mitigating feature degradation caused by model compression. A multi-scenario augmentation strategy is designed to balance convergence across different weather conditions. Experiments show that W-MTD’s student model improves detection accuracy by 7.0–13.9% under three adverse weather conditionscompared to the baseline teacher model trained solely on clear weather data while maintaining comparable clear-weather performance. With only 4 M parameters and 7 GFLOPs, the student model demonstrates favorable performance and efficiency compared to other real-time detectors, indicating its potential suitability for USV deployment. Full article

Background: The HEART score is a widely used risk-stratification tool in suspected acute coronary syndrome (ACS), but it still suffers from several limitations. We aim to assess its diagnostic accuracy for predicting coronary computed tomography angiography (CCTA) findings and explore possible enhancement by integrating additional clinical variables. Methods: In this retrospective, observational study, consecutive patients presenting to the Emergency Department with suspected ACS and undergoing CCTA were analyzed. The study assessed the HEART score’s diagnostic accuracy for predicting significant coronary artery stenosis (defined as ≥70% stenosis at CCTA) and explored improvements by integrating additional clinical variables for low-to-moderate-risk patients. Results: Three hundred seventy-nine patients were enrolled (age: 61 ± 15 years; male: 57%). According to the HEART score, 27% were at low risk, 67% moderate risk, and 6% high risk, with a prevalence of significant CAD of 7%, 27%, and 67%, respectively. The area under the curve (AUC) of the HEART score to predict significant CAD was 0.68. Male gender (OR = 1.76, 95% CI 1.03–3.02), right bundle branch block (OR = 4.15, 95% CI 1.66–10.40), and hemoglobin (OR = 1.21) and glucose levels (OR = 1.01) independently predicted significant coronary stenosis at CCTA in patients at low-to-moderate risk. Integrating these variables into the HEART score, the AUC improved from 0.68 to 0.74 (p = 0.004), with a net reclassification improvement of 13.5% (p = 0.032). Conclusions: Integrating additional clinical variables into the HEART score improves its accuracy to predict significant coronary artery stenosis at CCTA in suspected ACS patients at low-to-moderate risk. Tailoring assessments with these variables supports more accurate patient management and highlights the potential for more comprehensive diagnostic approaches. Full article

Accurate and rapid prediction of sensor responses is critical for real-time measurement systems, digital twin implementations, and sensor design optimization. Traditional numerical methods such as Finite Element Method (FEM) and Computational Fluid Dynamics (CFD) provide high-fidelity solutions but suffer from prohibitive computational costs, limiting their applicability in time-sensitive applications. This paper presents a novel framework utilizing Fourier Neural Operators (FNO) as surrogate models for fast multi-physics sensor response prediction across thermal, acoustic, and flow measurement domains. Unlike conventional neural networks that learn finite-dimensional mappings, FNO learns operators between infinite-dimensional function spaces by parameterizing the integral kernel in Fourier space, enabling resolution-invariant predictions with remarkable computational efficiency. We demonstrate the framework’s efficacy through three comprehensive case studies: (1) thermal sensor response prediction achieving $R^2>0.98$ with 8300× speedup over FEM, (2) acoustic sensor array modeling with mean absolute error below 0.5 dB and 4000× speedup over BEM, and (3) flow sensor characterization with velocity field prediction accuracy exceeding 97% and 31,000× speedup over CFD. The proposed FNO-based surrogate models are trained on simulation datasets generated from high-fidelity numerical solvers and validated against simulation holdout data for all three case studies, with additional experimental validation conducted for the thermal sensor case. Results indicate that FNO architectures effectively capture the underlying physics governing sensor behavior while reducing inference time from minutes to milliseconds. The framework enables real-time sensor calibration, uncertainty quantification, and design optimization, opening new possibilities for intelligent measurement systems and Industry 4.0 applications. We also investigate the spectral characteristics of FNO predictions, addressing the inherent low-frequency bias through a hybrid architecture combining FNO with local convolutional layers. The primary contributions of this work include: (1) the first systematic application of FNO-based surrogate modeling specifically tailored for sensor response prediction across multiple physics domains, (2) a novel H-FNO architecture that combines spectral operators with local convolutions to mitigate spectral bias in sensor applications, and (3) comprehensive validation including both simulation and experimental data for practical deployment. This work establishes FNO as a powerful tool for accelerating sensor simulation and advancing the field of AI-enhanced instrumentation and measurement. Full article

Polygon Boolean operations are widely used in integrated circuit (IC) layout processing tasks such as design rule checking (DRC) and optical proximity correction (OPC). Single-threaded Boolean algorithms cannot meet the efficiency demand of modern IC layouts, necessitating parallel algorithms for acceleration. However, existing parallel algorithms exhibit unsatisfactory parallel speedups and limited scalability, which typically stem from an inefficient merging phase that uses generic Boolean OR operations and redundantly reprocesses all edges of polygons on grid boundaries. To solve these problems, we proposed Polygon Tailor, a novel parallel algorithm for polygon Boolean operations that employs a data-parallel strategy and a new merging approach performing incremental XOR operations solely on edges along grid boundaries, eliminating redundant computations in previous methods. This innovation drastically reduces the grid-merging time by 1–2 orders of magnitude. Compared with the parallel implementation from a commercial layout processing tool, PolygonTailor is on average 5.08× faster and up to 14.36× faster for OR operations that generate highly complex polygons. Full article

Polymer composites and nanocomposites have become indispensable in aerospace, automotive, energy, electronics, soft robotics, and biomedical applications due to their high specific stiffness, strength, and manufacturability with highly tailorable multifunctional performance. Their rational design is complicated by strong, multiscale couplings among microstructural heterogeneity, interfacial physics, anisotropic response, and time- and temperature-dependent behavior, spanning molecular to structural length scales. This review provides a comprehensive survey of the principal computational methodologies used to predict and interpret the mechanical behavior of polymer composites and nanocomposites, highlighting the capabilities, specialties, and complementary roles of different modeling tools. This review first summarizes the essential physical characteristics governing polymer composites and nanocomposites. We then examine computational modeling approaches for polymer composites across four length scales: the constituent scale, microscale, mesoscale, and macroscale. For each scale, the primary modeling objectives, characteristic capabilities, and domains of applicability are discussed in the context of the existing literature. Cross-scale relationships and bridging strategies among these scales are also discussed, emphasizing how lower-scale simulations inform higher-scale models. The review then focuses on computational modeling of polymer nanocomposites, with particular attention to atomistic and coarse-grained molecular dynamics methods. Representative atomistic simulations, which capture interfacial structure, reinforcement–matrix interactions, and nanoscale mechanisms, are discussed. This is followed by discussions on coarse-grained approaches that extend the accessible length and time scales. Finally, we discuss how atomistic and coarse-grained models complement each other within integrated multiscale frameworks, enabling predictive links between nanoscale physics and macroscopic mechanical behaviors. Full article

The early detection of pulmonary nodules in chest CT scans is critical for improving lung cancer outcomes. While existing computer-aided diagnosis (CAD) systems have shown promise, most treat detection and segmentation as separate tasks, leading to fragmented pipelines and limited representation sharing. This study proposes a 2.5D multi-task learning (MTL) framework that integrates both tasks within a unified Mask R-CNN architecture. The framework incorporates a tailored preprocessing pipeline—including Hounsfield Unit (HU) normalisation, CLAHE enhancement, and lung parenchyma masking—to improve input consistency and task-relevant contrast characteristics. To enhance sensitivity for small or ambiguous nodules, an auxiliary RoI classifier is introduced. Additionally, a nodule-level evaluation strategy aggregates slice-wise predictions across the z-axis, supporting a clinically meaningful assessment that approximates 3D diagnostic workflows. Experiments on the LUNA16 dataset demonstrate that the proposed framework achieves a favourable trade-off between detection and segmentation performance under a unified 2.5D multi-task setting. These results highlight the potential of integrated MTL approaches to advance CAD systems for early lung cancer screening. Full article

Accurate and efficient ship detection in remote sensing imagery is a key enabler of intelligent maritime surveillance operations, supporting real-time decision-making in search and rescue, traffic management, and maritime law enforcement. However, remote ship images pose unique challenges for detection. These include densely distributed targets, complex sea-land backgrounds, large aspect ratios, diverse ship geometries, and high color similarity between ships and their surroundings. To address these issues under the computational constraints of unmanned aerial platforms, we propose GTO-YOLO11n, an enhanced YOLOv11n-based detection model tailored for efficient maritime ship sensing. First, we introduce the GatedFDConvBlock, which employs gated convolutional filtering to strengthen feature extraction for small and elongated ships while suppressing background clutter, thereby reducing missed and false detections in dense scenes. Second, we improve the C2PSA module with a dynamic multi-scale attention design, TSSABlock_DMS, to adaptively model cross-scale feature interactions and enhance robustness to complex maritime environments. Third, we replace the original detection head with OBB_ED, a parameter-sharing head that incorporates depthwise separable convolution (DSConv) and an angle prediction branch to lower model complexity while preserving high-quality localization and classification. To verify the performance of the algorithm, we were conducted on the public datasets HRSC2016, HRSC2016-MS, and ShipRSImageNet. The mAP@50 results were 95.2%, 88.3%, and 76.7%, showing improvements of 3.2%, 2.2%, and 2.6% compared to the original YOLOv11n. Full article

Texture image retrieval based on subjective visual descriptions remains a significant challenge due to the “semantic gap”, where conventional Content-Based Image Retrieval (CBIR) methods rely on low-level features or reference images that often diverge from human perception. To bridge this gap, this paper proposes a reference-free, perception-driven retrieval framework that enables users to query textures directly via abstract perceptual attributes. First, we constructed a human-centric perceptual feature space through controlled psychophysical experiments, quantifying 12 explicit texture attributes (e.g., granularity, directionality) using a 9-point Likert scale. Second, addressing the variability in visual sensitivity across user demographics, we developed a user-adaptive mechanism incorporating dual perceptual libraries tailored for art-major and non-art-major groups. Retrieval is formulated as a perception-aligned similarity optimization problem within this normalized space. Experimental evaluations on the Describable Textures Dataset (DTD) demonstrate that our method achieves superior perceptual consistency compared to both handcrafted descriptors (GLCM, LBP, HOG) and deep learning baselines (VGG16, ResNet50). Notably, the framework attained high PAP@3 performance across both user groups, validating its effectiveness in decoding fuzzy human intent without the need for query images. This work provides a robust solution for semantic-based texture retrieval in human–computer interaction scenarios. Full article

The emergence of the Smart Cities paradigm and the rapid expansion and integration of Internet of Things (IoT) technologies within this context have created unprecedented opportunities for high-resolution behavioral analytics, urban optimization, and context-aware services. However, this same proliferation intensifies privacy risks, particularly those arising from cross-modal data linkage across heterogeneous sensing platforms. To address these challenges, this paper introduces a comprehensive, statistically grounded framework for generating synthetic, multimodal IoT datasets tailored to Smart City research. The framework produces behaviorally plausible synthetic data suitable for preliminary privacy risk assessment and as a benchmark for future re-identification studies, as well as for evaluating algorithms in mobility modeling, urban informatics, and privacy-enhancing technologies. As part of our approach, we formalize probabilistic methods for synthesizing three heterogeneous and operationally relevant data streams—cellular mobility traces, payment terminal transaction logs, and Smart Retail nutrition records—capturing the behaviors of a large number of synthetically generated urban residents over a 12-week period. The framework integrates spatially explicit merchant selection using K-Dimensional (KD)-tree nearest-neighbor algorithms, temporally correlated anchor-based mobility simulation reflective of daily urban rhythms, and dietary-constraint filtering to preserve ecological validity in consumption patterns. In total, the system generates approximately 116 million mobility pings, 5.4 million transactions, and 1.9 million itemized purchases, yielding a reproducible benchmark for evaluating multimodal analytics, privacy-preserving computation, and secure IoT data-sharing protocols. To show the validity of this dataset, the underlying distributions of these residents were successfully validated against reported distributions in published research. We present preliminary uniqueness and cross-modal linkage indicators; comprehensive re-identification benchmarking against specific attack algorithms is planned as future work. This framework can be easily adapted to various scenarios of interest in Smart Cities and other IoT applications. By aligning methodological rigor with the operational needs of Smart City ecosystems, this work fills critical gaps in synthetic data generation for privacy-sensitive domains, including intelligent transportation systems, urban health informatics, and next-generation digital commerce infrastructures. Full article

This paper presents a hybrid deep learning framework for real-time sign language recognition (SLR) tailored to Internet of Things (IoT)-enabled environments, enhancing accessibility for Deaf communities. The proposed system integrates a Long Short-Term Memory (LSTM) network for static gesture recognition and a 3D Convolutional Neural Network (3D CNN) for dynamic gesture recognition. Implemented on a Raspberry Pi device using MediaPipe for landmark extraction, the system supports low-latency, on-device inference suitable for resource-constrained edge computing. Experimental results demonstrate that the LSTM model achieves its highest stability and performance for static signs at 1000 training epochs, yielding an average F1-score of 0.938 and an accuracy of 86.67%. In contrast, at 2000 epochs, the model exhibits a catastrophic performance collapse (F1-score of 0.088) due to overfitting and weight instability, highlighting the necessity of careful training regulation. Despite this, the overall system achieves consistently high classification performance under controlled conditions. In contrast, the 3D CNN component maintains robust and consistent performance across all evaluated training phases (500–2000 epochs), achieving up to 99.6% accuracy on dynamic signs. When deployed on a Raspberry Pi platform, the system achieves real-time performance with a frame rate of 12–15 FPS and an average inference latency of approximately 65 ms per frame. The hybrid architecture effectively balances recognition accuracy with computational efficiency by routing static gestures to the LSTM and dynamic gestures to the 3D CNN. This work presents a detailed epoch-wise comparative analysis of model stability and computational feasibility, contributing a practical and scalable IoT-enabled solution for inclusive, real-time sign-to-text communication in intelligent environments. Full article

This study employed an integrated experimental–computational methodology to investigate the critical role of the layer-stacking sequence in the acoustic performance of multi-layer porous materials for vehicle NVH applications. The acoustic properties of four distinct single-layer materials were first characterized via impedance tube measurements. A finite element simulation model based on the Johnson–Champoux–Allard (JCA) theory was subsequently developed in COMSOL Multiphysics 6.2 and rigorously validated. Leveraging this validated model, a systematic analysis was conducted on six different layer sequences under a fixed total thickness of 30 mm. The simulation results showed excellent agreement with experimental data, with a root-mean-square error (RMSE) below 5%. It was demonstrated that the stacking sequence significantly governed the mid-to-high frequency sound absorption behavior, which was strongly correlated with the modulation of the real and imaginary parts of the normalized surface acoustic impedance. This study thus demonstrated that the layer sequence—a previously underexplored design factor—critically determines the absorption performance of multi-layer materials at a fixed total thickness. A full design-space analysis revealed that performance shifts are governed by changes in interfacial acoustic impedance. This physics-driven insight provides a practical framework for tailoring absorbers to specific frequency bands, offering a viable path toward lightweight acoustic solutions for electric vehicle applications. Full article

With the rapid growth of e-commerce, intra-city express delivery has expanded rapidly, leading to various social issues, such as traffic congestion and air pollution. To address these problems, we focus on designing a multimodal intra-city express system in which parcels are collected from clients via local tours operated by a fleet of identical trucks, temporarily stored in satellite hubs, and then sent to the center hub via underground railway for further sorting and distribution. The problem involves capacitated hub location, client-to-hub allocation, and vehicle routing. Several practical constraints are considered in the routing aspect, including vehicle capacity, time windows, and maximum path length. With these practical considerations, we first formulate a star hub location-routing problem with time windows (SHLRPTW). Second, we use a branch-and-price-and-Benders-cut (BPBC) algorithm to solve it, which combines the Benders decomposition framework and branch-and-price-and-cut (BPC) framework. The BPBC algorithm is tailored, and several acceleration techniques are applied. Third, numerical experiments show that the proposed BPBC algorithm solves more instances and achieves smaller optimality gaps (0.75%) than CPLEX (19.55%) and the pure BPC algorithm (0.83%). The computational times are also critically reduced, with average speed-ups of 74.01 and 5.97, respectively. Furthermore, sensitivity analysis indicates that the BPBC algorithm performs much better than the BPC algorithm when the unit backbone transportation cost is high. Finally, case studies show the usefulness of the proposed model and algorithm. Full article

Background/Objectives: Type II endoleak (T2EL) remains the most frequent complication after endovascular aortic aneurysm repair (EVAR), with uncertain clinical relevance and management. While most resolve spontaneously, persistent T2ELs can lead to sac enlargement and rupture risk. This study proposes a deep learning framework for preoperative prediction of T2EL occurrence and severity using volumetric computed tomography angiography (CTA) data. Methods: A retrospective analysis of 277 patients undergoing standard EVAR (2010–2023) was performed. Preoperative CTA scans were processed for volumetric normalization and fed into a 3D convolutional neural network (CNN) trained to classify patients into three categories: no T2EL, benign T2EL, or malignant T2EL. The model was trained on 175 cases, validated on 72, and tested on an independent cohort of 30 patients. Performance metrics included accuracy, precision, recall, F1-score, and area under the receiver operating characteristic curve (AUC). Results: The CNN achieved an overall accuracy of 76.7% (95% CI: 0.63–0.90), a macro-averaged F1-score of 0.77, and an AUC of 0.93. Class-specific AUCs were 0.93 for no T2EL, 0.91 for benign, and 0.96 for malignant cases, confirming high discriminative capacity across outcomes. Most misclassifications occurred between adjacent categories. Conclusions: This study introduces the first end-to-end 3D CNN capable of predicting both the presence and severity of T2EL directly from preoperative CTA, without manual segmentation or handcrafted features. These findings suggest that preoperative imaging encodes latent structural information predictive of endoleak-driven sac reperfusion, potentially enabling personalized pre-emptive embolization strategies and tailored surveillance after EVAR. Full article

This review highlights the rapidly evolving role of artificial intelligence (AI) in transforming lung cancer care, with a specific focus on its integrated applications across diagnosis, biomarker discovery, and drug development. The novelty of this work lies in its holistic examination of how AI bridges these traditionally separate domains, from radiology and pathology to genomics and clinical trials, to create a more cohesive and personalized oncology pipeline. We detail how AI algorithms significantly enhance early detection by improving the accuracy and efficiency of pulmonary nodule characterization on computed tomography scans and enable precise cancer subtyping via computational pathology. In biomarker discovery, AI-driven analysis of radiomic features and genomic data facilitates the non-invasive prediction of tumor genotype, PD-L1 expression, and immunotherapy response, moving beyond invasive tissue biopsies. Furthermore, AI is accelerating the drug development lifecycle by identifying novel therapeutic targets and optimizing patient selection for clinical trials. The review also explores AI’s critical role in personalizing treatment regimens, including predicting outcomes for radiotherapy and immunotherapy, thereby tailoring therapy to individual patient profiles. We critically address the challenges of clinical translation, including model interpretability, data standardization, and ethical considerations, which are pivotal for real-world implementation. Finally, we contend that the future of lung cancer management hinges on robust, multi-institutional validation of AI tools and the development of trustworthy, explainable systems. Full article