Real-time Model Research Articles

YOLOv10-pose and YOLOv9-pose: Real-time strawberry stalk pose detection models

YOLOv10-pose and YOLOv9-pose: Real-time strawberry stalk pose detection models

An optimized lightweight real-time detection network model for IoT embedded devices.

With the rapid development of Internet of Things (IoT) technology, embedded devices in various computer vision scenarios can realize real-time target detection and recognition tasks, such as intelligent manufacturing, automatic driving, smart home, and so on. YOLOv8, as an advanced deep learning model in the field of target detection, has attracted much attention for its excellent detection speed, high precision, and multi-task processing capability. However, since IoT embedded devices typically own limited computing resources, direct deployment of YOLOv8 is a big challenge, especially for real-time detection tasks. To address this vital issue, this work proposes and deploys an optimized lightweight real-time detection network model that well-suits for IoT embedded devices, denoted as FRYOLO. To evaluate its performance, a case study based on real-time fresh and defective fruit detection in the production line is performed. Characterized by low training cost and high detection performance, this model accurately detects various types of fruits and their states, as the experimental results show that FRYOLO achieves 84.7% in recall and 89.0% in mean Average Precision (mAP), along with a precision of 92.5%. In addition, it provides a detection frame rate of up to 33 Frames Per Second (FPS), satisfying the real-time requirement. Finally, an intelligent production line system based on FRYOLO is implemented, which not only provides robust technical support for the efficient operation of fruit production processes but also demonstrates the availability of the proposed network model in practical IoT applications.

Advanced cloud-based real-time credit scoring models: Leveraging big data and AI

Cloud-based credit scoring systems represent a transformative shift in financial technology, leveraging artificial intelligence and machine learning to revolutionize creditworthiness assessment. These platforms integrate traditional and alternative data sources, enabling real-time processing and comprehensive risk evaluation. The evolution from conventional methods has enhanced accessibility for underserved populations while maintaining robust security measures and regulatory compliance. Through advanced technical architecture incorporating data collection, processing, and analysis layers, these systems deliver faster, more accurate credit decisions. The implementation of sophisticated fraud detection mechanisms, coupled with machine learning algorithms, has significantly reduced false positives and improved risk assessment capabilities. Industry leaders like TransUnion demonstrate the practical success of these innovations through improved operational efficiency and expanded market reach. Despite challenges in data quality management and regulatory compliance, the future of credit scoring continues to advance with blockchain integration, enhanced real-time processing, and innovative fraud detection capabilities.

Big data and impact in demand-driven material requirements planning

Big Data integrated with DDMRP has turned a new leaf in managing the supply chain. The research elucidates the deep impact of Big Data in relation to DDMRP. Big Data from online transactions, sensor data, social interactions, and many other areas possesses the potential to bring a paradigm shift in the area of demand forecasting, inventory management, and supply chain optimization as a whole. It allows organizations to make informed, agile decisions for cost reduction, inventory optimization, and improved customer service by enabling real-time data analysis and predictive modelling. However, there are challenges to the adoption of Big Data in DDMRP, including data security and advanced analytics capability. The study examines the application, benefits, challenges, and impact of Big Data implementation in DDMRP. Therefore, the study also provides valuable input for organizations interested in making use of the potential of big data. Its findings point toward a decisive role of big data in reshaping supply chain strategy and enhancing responsiveness to market fluctuations for modern business.

Integrating Python and MATLAB for Digital Twin Applications in Aeroservoelasticity

The development of Digital Twin (DT) applications in intricate engineering areas has been greatly accelerated by the incorporation of sophisticated computational tools like Python and MATLAB. The goal of this research is to model, simulate, and analyze aeroservoelastic systems for Digital Twin implementations by utilizing the combined powers of Python and MATLAB. In order to achieve the real-time predictive skills needed for Digital Twins, aeroservoelasticity—which involves the coupled interplay of aerodynamic, structural, and control forces—presents special obstacles. This work successfully tackles these issues by fusing MATLAB's strong numerical and control system capabilities with Python's adaptability and library ecosystem. Using real-time sensor data streaming, and feedback control system implementation is used in this work. High-fidelity state-space modeling, sophisticated control design (such as PID and state feedback controllers), and system dynamics visualization are all done concurrently with MATLAB. In order to simulate and stabilize a highly flexible aeroelastic system, a continuous feedback loop was modeled in both environments, highlighting the platforms' complimentary advantages. Through the MATLAB Engine API for Python, the integrated method facilitates smooth data transmission between Python and MATLAB, guaranteeing real-time communication and model parameter synchronization. The efficiency of this dual-platform system is demonstrated by case studies on wing flutter suppression, limit cycle oscillation (LCO) mitigation, and control input optimization. The method is well suited for the dynamic needs of aeroservoelastic Digital Twins because of the results, which demonstrate a shorter error convergence time, improved stability, and computational economy. By utilizing MATLAB and Python's interoperability, this work establishes the groundwork for scalable, real-time Digital Twin solutions in aeronautical engineering. It highlights how crucial hybrid computational ecosystems are for bridging the gap between high- fidelity simulations and real-time predictive capabilities, opening the door for improvements in fault prediction, control optimization, and aeronautical system monitoring.

ESF-DETR: a real-time and high-precision detection model for cigarette appearance

ESF-DETR: a real-time and high-precision detection model for cigarette appearance

Prediction model and real-time diagnostics of hydraulic fracturing pressure for highly deviated wells in deep oil and gas reservoirs

It is a common occurrence in the fracture processes of deep carbonate reservoirs that the fracturing construction pressure during hydraulic fracturing operation exceeds 80 MPa. The maximum pumping pressure is determined by the rated pressure of the pumping pipe equipment and the reservoir characteristics, which confine the fracture to the target area. When the pump pressure exceeds the safety limit, hydraulic fracturing has to reduce the construction displacement to prevent potential accidents caused by overpressure. Therefore, real-time prediction of the fracturing construction pressure and diagnosis of abnormal fluctuations during hydraulic fracturing of highly deviated wells are indispensable. Based on the well trajectory, pumping process, and string structure of highly deviated wells, a movement interface model for the fracturing fluid at different stages within the wellbore has been established, using the method of computational fluid dynamics. This model analyzes the fluid movement behavior with diverse properties at various fracturing times and determines the relationship between the pressure changes at the leading and trailing edges of fluid movement in each section of the wellbore over time by combining different string structures and preset pumping procedures. The frictional pressure within the wellbore fluid, the hydrostatic fluid pressure, and the near-well friction drags have been calculated and predicted. A real-time prediction model for diagnosing pumping fracturing has been constructed to further comprehend “abnormal” fracturing construction pressures in highly deviated wells. This offers a theoretical foundation for the correct diagnosis and decision-making regarding hydraulic fracturing in highly deviated wells while guiding its smooth implementation in real time.

Socially interactive industrial robots: a PAD model of flow for emotional co-regulation

This article presents the development of a socially interactive industrial robot. An Avatar is used to embody a cobot for collaborative industrial assembly tasks. The embodied covatar (cobot plus its avatar) is introduced to support Flow experiences through co-regulation, interactive emotion regulation guidance. A real-time continuous emotional modeling method and an aligned transparent behavioral model, BASSF (Boredom, Anxiety, Self-efficacy, Self-compassion, Flow) is developed. The BASSF model anticipates and co-regulates counterproductive emotional experiences of operators working under stress with cobots on tedious industrial tasks. The targeted Flow experience is represented in the three-dimensional Pleasure, Arousal, and Dominance (PAD) space. We present how, despite their noisy nature, PAD signals can be used to drive the BASSF model with its theory-based interventions. The empirical results and analysis provides empirical support for the theoretically defined model, and clearly points to the need for data pre-filtering and per-user calibration. The proposed post-processing method helps quantify the parameters needed to control the frequency of intervention of the agent; still leaving the experimenter with a run-time adjustable global control of its sensitivity. A controlled empirical study (Study 1, N = 20), tested the model’s main theoretical assumptions about Flow, Dominance, Self-Efficacy, and boredom, to legitimate its implementation in this context. Participants worked on a task for an hour, assembling pieces in collaboration with the covatar. After the task, participants completed questionnaires on Flow, their affective experience, and Self-Efficacy, and they were interviewed to understand their emotions and regulation during the task. The results from Study 1 suggest that the Dominance dimension plays a vital role in task-related settings as it predicts the participants’ Self-Efficacy and Flow. However, the relationship between Flow, pleasure, and arousal requires further investigation. Qualitative interview analysis revealed that participants regulated negative emotions, like boredom, also without support, but some strategies could negatively impact wellbeing and productivity, which aligns with theory. Additional results from a first evaluation of the overall system (Study 2, N = 12) align with these findings and provide support for the use of socially interactive industrial robots to support wellbeing, job satisfaction, and involvement, while reducing unproductive emotional experiences and their regulation.

NoctuDroneNet: Real-Time Semantic Segmentation of Nighttime UAV Imagery in Complex Environments

Nighttime semantic segmentation represents a challenging frontier in computer vision, made particularly difficult by severe low-light conditions, pronounced noise, and complex illumination patterns. These challenges intensify when dealing with Unmanned Aerial Vehicle (UAV) imagery, where varying camera angles and altitudes compound the difficulty. In this paper, we introduce NoctuDroneNet (Nocturnal UAV Drone Network, hereinafter referred to as NoctuDroneNet), a real-time segmentation model tailored specifically for nighttime UAV scenarios. Our approach integrates convolution-based global reasoning with training-only semantic alignment modules to effectively handle diverse and extreme nighttime conditions. We construct a new dataset, NUI-Night, focusing on low-illumination UAV scenes to rigorously evaluate performance under conditions rarely represented in standard benchmarks. Beyond NUI-Night, we assess NoctuDroneNet on the Varied Drone Dataset (VDD), a normal-illumination UAV dataset, demonstrating the model’s robustness and adaptability to varying flight domains despite the lack of large-scale low-light UAV benchmarks. Furthermore, evaluations on the Night-City dataset confirm its scalability and applicability to complex nighttime urban environments. NoctuDroneNet achieves state-of-the-art performance on NUI-Night, surpassing strong real-time baselines in both segmentation accuracy and speed. Qualitative analyses highlight its resilience to under-/over-exposure and small-object detection, underscoring its potential for real-world applications like UAV emergency landings under minimal illumination.

TLSynth: A Novel Blender Add-On for Real-Time Point Cloud Generation from 3D Models

Point clouds are a crucial element in the process of scanning and reconstructing 3D environments, such as buildings or heritage sites. They allow for the creation of 3D models that can be used in a wide range of applications. In some cases, however, only the 3D model of an environment is available, and it is necessary to obtain point clouds with the same characteristics as those captured by a laser scanner. For instance, point clouds may be required for surveys, performance optimization, site scan planning, or validation of point cloud processing algorithms. This paper presents a new terrestrial laser scanner (TLS) simulator, designed as a Blender add-on, that produces synthetic point clouds from 3D models in real time. The simulator allows users to adjust a set of parameters to replicate real-world scanning conditions, such as noise generation, ensuring the synthetic point clouds closely mirror those produced by actual laser scanners. The target meshes may be derived from either a real-world scan or 3D designs created using design software. By replicating the spatial distributions and attributes of real laser scanner outputs and supporting real-time generation, the simulator serves as a valuable tool for scan planning and the development of synthetic point cloud repositories, advancing research and practical applications in 3D computer vision.

SP-YOLO: A Real-Time and Efficient Multi-Scale Model for Pest Detection in Sugar Beet Fields.

Beet crops are highly vulnerable to pest infestations throughout their growth cycle, which significantly affects crop development and yield. Timely and accurate pest identification is crucial for implementing effective control measures. Current pest detection tasks face two primary challenges: first, pests frequently blend into their environment due to similar colors, making it difficult to capture distinguishing features in the field; second, pest images exhibit scale variations under different viewing angles, lighting conditions, and distances, which complicates the detection process. This study constructed the BeetPest dataset, a multi-scale pest dataset for beets in complex backgrounds, and proposed the SP-YOLO model, which is an improved real-time detection model based on YOLO11. The model integrates a CNN and transformer (CAT) into the backbone network to capture global features. The lightweight depthwise separable convolution block (DSCB) module is designed to extract multi-scale features and enlarge the receptive field. The neck utilizes the cross-layer path aggregation network (CLPAN) module, further merging low-level and high-level features. SP-YOLO effectively differentiates between the background and target, excelling in handling scale variations in pest images. In comparison with the original YOLO11 model, SP-YOLO shows a 4.9% improvement in mean average precision (mAP@50), a 9.9% increase in precision, and a 1.3% rise in average recall. Furthermore, SP-YOLO achieves a detection speed of 136 frames per second (FPS), meeting real-time pest detection requirements. The model demonstrates remarkable robustness on other pest datasets while maintaining a manageable parameter size and computational complexity suitable for edge devices.

Camouflaged Object Detection System

The detection of camouflaged objects is crucial for applications in surveillance, wildlife monitoring, and military scenarios, where objects blend seamlessly into their surroundings. This review consolidates 15 influential research studies covering advancements in datasets, models, real-time detection technologies, and multimodal approaches. The focus is on implementing YOLOv8, a state-of-the-art real-time object detection model, using the ACD1K dataset, which is specifically designed for military surveillance. By synthesizing methodologies, evaluation metrics, and applications, this paper highlights significant progress in camouflaged object detection (COD) and identifies ongoing challenges in computational efficiency, dataset diversity, and real-world adaptability. Key Words: Camouflaged object detection, YOLOv8, ACD1K dataset, military surveillance, real-time detection.

Intelligent Bayesian Neural Networks for Stochastic SVIS Epidemic Dynamics: Vaccination Strategies and Prevalence Fractions with Wiener Process

Real-time forecasting of infectious diseases is crucial for effective public health management, particularly during outbreaks. When infectious disease predictions are based on mechanistic models, they can guide resource allocation and help evaluate the potential effects of different interventions. However, accurately parametrizing these models in real time presents a challenge, as timely information on behavioral shifts, interventions and transmission pathways is often lacking. This investigation leverages the artificial neural networks with the Bayesian regularization (BR-ANN) backpropagation approach to examine the dynamical pathogen spread with Wiener process incorporation. The stochastic differential model is structured into susceptible, vaccinated, infectious and susceptible (SVIS) compartments. The Kloeden–Platen–Schurz (KPS) computing paradigm for the stochastic differential system is utilized to generate synthetic datasets by applying transformations to key factors, including the population recruitment ratio, transmission ratio of vulnerable individuals, natural death rate of the population, vaccination rate of vulnerable individuals, total population size, immune loss ratio of susceptible individuals, recovery rate and mortality rate from the disease among infected individuals. Random selection from the generated datasets is exploited for the training and testing procedures for constructing the BR-ANN networks. The significance of the proposed scheme for various stochastic SVIS system scenarios is endorsed by the comprehensive assessments of the BR-ANN approach that are conducted by means of extensive experimentations and comparison with the reference KPS solutions of the SVIS system in terms of MSE optimal performance plots, absolute errors, autocorrelation analysis, regression indices and error histograms.

Dynamic Simulation of Photothermal Environment in Solar Greenhouse Based on COMSOL Multiple Physical Fields

Solar greenhouses are essential facilities for agricultural production in northern China, where uneven internal environments pose significant challenges. This study established a numerical model of photothermal conditions in solar greenhouses. Utilizing COMSOL MultiphysicsTM, we established a microclimate model that encompasses the greenhouse exterior and the soil directly below it, without considering the crops. This model coupled multiphysical fields with fluid flow and heat transfer processes. The boundary conditions and initial values of the external environment and soil were derived from meteorological data and an efficient interpolation function method, with the time step updated every 1h. The results demonstrate that the simulated values were in good agreement with the measured values. Our findings reveal the temporal dynamics of radiation and temperature changes, as well as spatial heterogeneity, within solar greenhouses under different winter weather conditions. Additionally, the potential of integrating with other real-time monitoring and control models was discussed. This study provides a theoretical foundation for developing microclimate models and predicting photothermal environments in greenhouses.

Real-Time Models for Manufacturing Processes: How to Build Predictive Reduced Models

Real-Time Models for Manufacturing Processes: How to Build Predictive Reduced Models

A laser tracking attitude dynamic measurement method based on real-time calibration and AEKF with a multi-factor influence analysis model

Abstract To address the limitations of six-degree-of-freedom (6-DOF) visual tracking systems in terms of low dynamic performance and the inability to perform measurements under light occlusion, this paper proposes a novel laser tracking attitude dynamic measurement method integrating vision and inertial measurement unit (IMU). A tailored real-time calibration model is developed to enable precise alignment of the vision and IMU coordinate systems during dynamic operations. To overcome the limitations of extended Kalman filter (EKF) that rely on static noise assumptions in complex dynamic measurement scenarios, this study proposes an adaptive mechanism based on multi-factor influence analysis. Through dynamic monitoring of visual observation noise, an adaptive EKF algorithm is designed to enhance performance under varying conditions. When integrated into a vision/IMU fusion system, the algorithm significantly enhances the system’s sensitivity to variations in visual sensor observation noise caused by changes in lighting and motion states. This capability allows rapid adjustment of filtering parameters, leading to substantial improvements in filtering accuracy and stability. Simulation results demonstrate that the proposed algorithm outperforms traditional EKF in fusion measurement accuracy when tested in simulated noise environments featuring sinusoidal and random jump signals. Furthermore, the fusion measurement accuracy is mainly influenced by the data density of the visual sensor and the measurement distance. Finally, the proposed method was validated on a precision turntable. Experimental results reveal that at a 3 m measurement distance and within a 0°–25° angular range, when the turntable rotates at a constant speed of 5°/s along a single axis, the maximum absolute deviation in repeatability of the fused attitude measurements was below 0.11°, and the system is capable of achieving a high measurement frequency of 100 Hz.

YOLOv7-CWFD for real time detection of bolt defects on transmission lines

Detecting bolt defects on transmission lines is crucial for ensuring the safe operation of the electrical power system. However, existing methods for detecting bolt defects on transmission lines require higher detection accuracy and smaller model sizes. To address these challenges, this paper proposes a real-time bolt defect detection model based on YOLOv7, named YOLOv7-CWFD. The model integrates the Channel Shuffle Diverse Path Aggregation Network (CSDPAN), significantly reducing computational and parameter complexity while maintaining high detection accuracy. Additionally, weighted Efficient Intersection over Union (EIoU) and Normalized Wasserstein Distance (NWD) loss functions are designed to reduce the network’s sensitivity to object size variations and enhance model convergence in regression tasks. The Fast Fourier Channel Attention Mechanism (FFCAM) is introduced between the backbone and neck fusion networks to mitigate excessive smoothing of detailed information and improve the network’s sensitivity to objects. The DySample upsampling operator is implemented to replace the upsampling module in the neck fusion network, minimizing information loss during the upsampling process. Experiments conducted on the custom Transmission Line Bolt Defect Dataset (TLBDD) demonstrate a reduction of 10.30MB in model parameter size, along with a 2.30% increase in mean Average Precision (mAP) compared with the original YOLOV7 and a detection speed of 51.15 frames per second (FPS). Experiments on the public dataset CCTSDB further confirm the model’s robust generalization capability. These experiments validate the effectiveness of the proposed algorithm.

Uncovering Dynamical Equations of Stochastic Decision Models Using Data-Driven SINDy Algorithm.

Decision formation in perceptual decision making involves sensory evidence accumulation instantiated by the temporal integration of an internal decision variable toward some decision criterion or threshold, as described by sequential sampling theoretical models. The decision variable can be represented in the form of experimentally observable neural activities. Hence, elucidating the appropriate theoretical model becomes crucial to understanding the mechanisms underlying perceptual decision formation. Existing computational methods are limited to either fitting of choice behavioral data or linear model estimation from neural activity data. In this work, we made use of sparse identification of nonlinear dynamics (SINDy), a data-driven approach, to elucidate the deterministic linear and nonlinear components of often-used stochastic decision models within reaction time task paradigms. Based on the simulated decision variable activities of the models and assuming the noise coefficient term is known beforehand, SINDy, enhanced with approaches using multiple trials, could readily estimate the deterministic terms in the dynamical equations, choice accuracy, and decision time of the models across a range of signal-to-noise ratio values. In particular, SINDy performed the best using the more memory-intensive multitrial approach while trial averaging of parameters performed more moderately. The single-trial approach, although expectedly not performing as well, may be useful for real-time modeling. Taken together, our work offers alternative approaches for SINDy to uncover the dynamics in perceptual decision making and, more generally, for first-passage time problems.

Advancing Building and Construction Higher Education: The Online Real-Time Block Model’s Contributions to Professional Skills, Gender Equity, and Industry Preparedness

Traditional In-Person Semester-Length (IP-SL) courses often struggle with inherent time constraints, lack of flexibility, and geographic limitations, delaying effective learning and accessibility for students. Moreover, the extended duration of the Semester-Length (SL) structure reduce focus due to engagement with multiple subjects simultaneously, increased stress, and limited timely feedback and assessment. This study evaluates the Online Real-Time Block Model (ORT-BM), an intensive online model, highlighting its potential to enhance engagement, satisfaction, and inclusivity in project-based programs like construction in higher education. Building surveying as a critical field in construction is selected as the case study since professional surveyors must stay current with rapidly evolving building codes, regulations, and sustainability practices. However, the rigid structure of IP-SL courses often leaves graduates less prepared to meet industry needs. Conducting a comparative analysis of a case study, the Bachelor of Building Surveying program (NBBS) at Victoria University, the research compares three teaching models: IP-SL (2016–2018), In-Person Block Model (IP-BM, 2019–2020), and ORT-BM (2020–2023) using Student Evaluation of Units (SEU) data and Quality Indicators for Learning and Teaching (QILT) metrics. Findings, derived from SEU and QILT, reveal that ORT-BM improves student satisfaction, accelerates course completion rates, and fosters gender equity through inclusive learning environments while enhancing accessibility for geographically dispersed and disadvantaged students. By integrating advanced digital tools like virtual site visits, ORT-BM enhances professional readiness, aligning education with evolving industry standards. Future research may explore developing hybrid models to optimize cognitive load further, improve accessibility, and enhance flexibility.

Machine Learning Models for the Early Real-Time Prediction of Deterioration in Intensive Care Units-A Novel Approach to the Early Identification of High-Risk Patients.

Background Predictive machine learning models have made use of a variety of scoring systems to identify clinical deterioration in ICU patients. However, most of these scores include variables that are dependent on medical staff examining the patient. We present the development of a real-time prediction model using clinical variables that are digital and automatically generated for the early detection of patients at risk of deterioration. Methods Routine monitoring data were used in this analysis. ICU patients with at least 24 h of vital sign recordings were included. Deterioration was defined as qSOFA ≥ 2. Model development and validation were performed internally by splitting the cohort into training and test datasets and validating the results on the test dataset. Five different models were trained, tested, and compared against each other. The models were an artificial neural network (ANN), a random forest (RF), a support vector machine (SVM), a linear discriminant analysis (LDA), and a logistic regression (LR). Results In total, 7156 ICU patients were screened for inclusion in the study, which resulted in models trained from a total of 28,348 longitudinal measurements. The artificial neural network showed a superior predictive performance for deterioration, with an area under the curve of 0.81 over 0.78 (RF), 0.78 (SVM), 0.77 (LDA), and 0.76 (LR), by using only four vital parameters. The sensitivity was higher than the specificity for the artificial neural network. Conclusions The artificial neural network, only using four automatically recorded vital signs, was best able to predict deterioration, 10 h before documentation in clinical records. This real-time prediction model has the potential to flag at-risk patients to the healthcare providers treating them, for closer monitoring and further investigation.