课程文献中心 更多>>

关键词： Aroma quality monitoring   Interpretable CNN   Odor sensor   Tea processing   Visible near-infrared spectroscopy  

摘要： The accurate prediction of the withering degree is essential for ensuring the aroma quality of black tea. In this study, we present an innovative intelligent platform that integrates a carefully designed colorimetric sensor array (CSA) with visible near-infrared (VNIR) spectroscopy, powered by an interpretable deep learning model (convolutional neural network, CNN) to predict withering degrees. Specifically, eight dyes were selected from sixteen candidates based on correlation analysis (R² > 0.36). These dyes were drop-cast onto filter paper to create the CSA, which was used to capture key volatile organic compounds (VOCs) released during the withering of tea leaves. Spectral data from the CSA were collected and modeled to identify the different withering stages. To further enhance model interpretability, gradient-weighted class activation mapping (Grad-CAM) was applied to highlight the spectral regions most relevant to classification. When compared to conventional models, the Grad-CAM-CNN demonstrated superior performance, achieving 99% accuracy and a computational efficiency of 0.0559 s/100 samples. This approach not only ensures the aroma quality of black tea but also provides an efficient, scalable, and cost-effective solution for real-time process control in tea manufacturing, with potential for adaptation to other food systems.

关键词： Adaptive substepping explicit schemes   Consistent tangent operator   Elasto-plastic analysis   Runge–Kutta method  

摘要： Elasto-plastic finite element analysis relies on constitutive integration at the material-level and nonlinear iteration at the global-level. Adaptive substepping explicit integration schemes exhibit high accuracy, efficiency, and robustness in constitutive computations, but have long suffered from the lack of a rigorous consistent tangent operator (CTO), which is essential for achieving quadratic convergence of Newton–Raphson iterations. This study presents a systematic derivation of the CTO for adaptive substepping explicit Runge–Kutta methods, accounting for its evolution during yield surface intersection, adaptive substepping, and drift correction, thereby enabling fully algorithmic linearization. An explicit recursive formulation is derived to update the CTO throughout the substepping process. For the Modified Cam Clay model, an improved scheme for computing the yield surface intersection point is proposed, significantly enhancing computational efficiency. Numerical results from element tests and finite element simulations demonstrate that the proposed method achieves global quadratic convergence.

摘要： Coronary artery disease (CAD) remains the leading cause of mortality in the United States, and despite decades of innovation in secondary prevention, recurrent cardiovascular events continue to occur, amassing an enormous clinical and economic bur...

关键词： Sensor fusion   SLAM   LiDAR odometry  

摘要： High-altitude downward-looking LiDAR odometry remains challenging due to point cloud sparsity, scale variation, and geometric degeneration. This paper presents NDT-CAM, a visual-assisted LiDAR odometry framework for high-altitude UAV surveying and mapping. The proposed method integrates segmented motion compensation, a scale perceptron, and visual constraints within an NDT-based registration framework. Segmented motion compensation is used to alleviate motion distortion under dynamic flight, while the scale perceptron adaptively adjusts the voxel size and local search range according to the spatial extent of the current scene. To improve robustness in geometrically weak regions, visual feature information is incorporated as an auxiliary consistency constraint into the iterative optimization. Experiments on both public and private datasets demonstrate that the proposed method achieves robust and competitive localization performance across different flight altitudes, LiDAR types, and platform settings. The method maintains real-time performance over an operating range covering altitudes from 0 m to 130 m and speeds from 0 m/s to 12 m/s. These results suggest that adaptive, visual-assisted NDT provides a practical solution for high-altitude UAV mapping under weak GNSS/IMU dependence. We open-source the code and dataset on https://***/huashu996/NDT-CAM.

关键词： Computer-aided design   CAD generative modeling   Image-guided modeling   Multimodal learning   Interpretable modeling  

摘要：

• •
  This paper introduces an interpretable framework for CAD generation and editing.
• •
  We proposes a modular architecture with CoT for human-centric design.
• •
  We build a new dataset, Mind-CAD EditBench, for instruction-driven CAD editing.

关键词： Abnormality   Artificial intelligence   Boosting   CAD   Deciduous   Deep learning   Dental   DNN models   Ensemble learning   Furcation detection   Image processing   Intraoral periapical radiographs   Mandible   Object detection   Radiographs   Segmentation  

摘要： In dentistry, Artificial Intelligence (AI) plays a crucial role in transforming the process of disease detection and diagnosis. Among the numerous applications, AI is particularly valuable in the early detection of dental problems, like furcation involvement. This problem is especially common in children’s teeth, can penetrate deep into the tooth structure and potentially affect the developing permanent teeth. If not treated promptly, it may lead to serious problems, including tooth root bifurcation and eventual tooth loss. To tackle this problem, the present research introduces FD-CAD (Furcation Detection Computer-Aided Diagnosis), a state-of-the-art Computer-Aided Diagnosis system developed to detect furcation involvement in radiographs of deciduous mandibular molars. This FD-CAD system leverages advanced Deep learning (DL) models optimizing the Intraoral periapical radiographs (IOPA), including You Look Only Once (YOLO) models, Faster region-based convolutional neural network (Faster-RCNN), Fully convolutional one-stage object detector (FCOS), and RetinaNet. A comprehensive performance evaluation of these models was conducted, and Faster-RCNN emerged as the top contender. Furthermore, the diagnostic capability of the system was significantly enhanced using a domain-adapted boosting-based approach, Deciduous detection boosting (DDB). Moreover, this enhancement led to a significant improvement in performance, achieving an Average precision (AP) of 96.62% at an IoU threshold of 0.5 (AP50) and 95.39% at an Intersection over Union (IoU) of 0.7 (AP70). Additionally, these results demonstrate the reliability and robustness of the proposed method in detecting furcation involvement in deciduous teeth. In conclusion, these findings highlight the potential of DL in transforming early dental diagnosis, particularly in pediatric dentistry, and set the stage for more effective and preventive dental care.

关键词： CAM   Co-training   Dual-branch   Pseudo-label   Weakly supervised semantic segmentation  

摘要： Weakly Supervised Semantic Segmentation (WSSS) remains highly challenging because image-level supervision typically produces class activation maps (CAMs) that are incomplete or noisy when used as pixel-level pseudo-labels. Despite the architectural efficiency of one-stage approaches, they are often hindered by tight encoder-label coupling: CAMs and segmentation predictions are derived from the same encoder and optimized jointly, leading to the propagation and reinforcement of initial CAM inaccuracies by the segmentation outputs. To circumvent this limitation, we propose SynerNet, a one-stage dual-branch framework that explicitly mandates complementary yet synergistic objectives: one branch generates broad pseudo-labels to enhance coverage, while the other produces precise pseudo-labels to sharpen localization. With such pseudo-labels, the segmentation network yields simultaneously comprehensive and accurate predictions. The broad branch (B-CAM) leverages global attention to expand foreground coverage by guiding ambiguous regions toward likely foreground, whereas the precise branch (P-CAM) emphasizes fine localization by encouraging unreliable pixels toward the background. Through cross-supervision, the two branches effectively decouple the optimization process, alleviating the risk of error reinforcement inherent in direct coupling. To further integrate their strengths, we introduce a confidence matrix derived from multi-scale ViT features, in which pixels consistently classified across layers are treated as high-confidence, while inconsistent ones are marked as uncertain. This enables a confidence-guided fusion strategy that directly adopts reliable predictions and adaptively blends uncertain regions using contributions from both branches. Such a complementary design mitigates error reinforcement and promotes mutually beneficial learning, enabling the network to generate high-fidelity pseudo-labels in a fully end-t

关键词： Nanotoxicology   Drug delivery systems (DDS)   Advanced in vitro models   Advanced in vivo models   Non-mammalian models   2D   two-dimensional   2,3-Bis-(2-methoxy-4-nitro-5-sulfophenyl)−2H-tetrazolium-5-carboxanilide   XTT   2,4-dinitrophenylhydrazine   DNPH   3-(4,5-dimethylthiazol-2-yl)−2,5-diphenyltetrazolium bromide   MTT assay   AgNPs   silver nanoparticles   AI   artificial intelligence   ALI   air–liquid interface   AuNPs   gold nanoparticles   BBB   blood–brain barrier   CAM   chorioallantoic membrane   CdO-NPs   cadmium oxide nanoparticles   CrO3-NPs   chromium trioxide nanoparticles   Cu-NPs   copper nanoparticles   CuO-NPs   copper oxide nanoparticles   DDS   drug delivery systems   ECM   extracellular matrix   EBB   extrusion-based bioprinting   ESCs   embryonic stem cells   Fe2O3, Fe3O4-NPs   iron oxide nanoparticles   HTS   high-throughput screening   iPSCs   induced pluripotent stem cells   LDH   lactate dehydrogenase   MALIES   modular air–liquid interface exposure system   MCTSs   multicellular tumor spheroids   ML   machine learning   NAMs   new approach methodologies   NEs   nanoemulsions   NPs   nanoparticles   OoC   organ-on-chip   PCTS   precision-cut tissue slices   PD   pharmacodynamic   PE-NPs   polyethylene nanoparticles   PK   pharmacokinetic   PSNPs   polystyrene nanoplastics   QSARs   quantitative structure–activity relationships   QSPRs   quantitative structure–property relationships   SiO2   silica nanoparticles   TiO2-NPs   titanium dioxide nanoparticles   ZnO-NPs   zinc oxide nanoparticles  

摘要： Nanomaterials have been widely used across medical and health sciences due to their unique physicochemical characteristics, versatile functionalisation, and their remarkable capacity to interact with (and often penetrate) biological tissues at the molecular and cellular level. These properties have made nanomaterials particularly attractive for drug delivery systems, improving targeted delivery and therapeutic efficacy. Nevertheless, the same features that make nanomaterials so valuable in biomedical contexts may also give rise to unforeseen biological interactions, including oxidative stress, inflammation, immune responses, and organ accumulation, thereby raising legitimate concerns about their potential toxicological implications.

关键词： Constructive solid geometry   CAD models   3D generation   Large language models  

摘要：

• •
  Two-stage CSG framework for interpretable CAD generation from natural language.
• •
  Structure-aware Transformer with hierarchical supervision and validity loss.
• •
  Large-scale CSG dataset with paired text for text-driven CAD modeling.

关键词： Mid-surface abstraction   Variable thin-walled models   GPU acceleration  

摘要：

• •
  End-to-end GPU acceleration for the most critical aspects of Mid-surface abstraction.
• •
  Full-GPU face pairing with hierarchical culling and fast pruning criteria.
• •
  Branchless mid-point solver for variable thickness with a bracket-and-bisect kernel.
• •
  Extensive evaluation of diverse models from different real-world domains.