To address the challenges of a high missed detection rate for small targets and strong interference from complex backgrounds in remote sensing image target detection, the paper proposes an improved YOLOv11n based method. We introduce an enhanced YOLOv11n model incorporating a dynamic receptive field module (RFAConv) and a snake deformation modeling module (DySnakeConv). This approach strengthens shallow feature extraction capabilities and refines adaptive fitting of target boundaries, thereby improving detection accuracy. Experimental results demonstrate that on the RSOD dataset, the improved model achieves mean average precision (mAP) scores of 96.9 % at IoU = 0.50 (mAP50) and 65.5 % over IoU thresholds from 0.50 to 0.95 (mAP5095). These results surpass those of YOLOv8n, YOLOv10n, and other comparative models in key metrics such as precision and recall. Importantly, the model maintains comparable performance on the NWPU VHR-10 dataset. The proposed model presents an efficient solution for detecting small and geometrically sensitive targets in high-resolution remote sensing images.

Modern high-performance sports place increasing demands on athletes’ physical, technical, and psychological preparedness, intensifying the challenge of sports injuries and overtraining. Traditional monitoring methods often lack predictive precision, hindering timely identification of injury risks.This study develops and compares three LSTM-based models for predicting injury risk in runners: one leveraging biomechanical parameters, another using psychophysiological indicators, and an integrated model combining both. Models were developed using data from digital twins of two professional runners, incorporating physiological (heart rate, heart rate variability, lactate levels), biomechanical (joint angles, step symmetry, accelerations), and psychophysiological (sleep quality, fatigue, cognitive responses) metrics. The integrated model demonstrated superior performance, achieving an Accuracy of 0.89, F1-score of 0.87, and AUC-ROC of 0.91. SHAP analysis identified key predictors, including step symmetry, tibial shock, reduced heart rate variability, sleep quality decline, and subjective fatigue. These findings highlight the enhanced predictive power of integrating diverse data types, offering a robust foundation for personalized injury prevention systems in sports.

Effective ICT governance is essential in the public sector to drive digital transformation and improve service delivery. This research investigates the corporate governance of ICT Policy Framework (CGICTPF) and Public Finance Management Act (PFMA) and State Information Technology Agency (SITA) Act governs the operational activities and strategic directions of Government Information Technology Officers (GITOs) in Eastern Cape, KwaZulu-Natal and Free State provincial administrations in South Africa. Using a comparative case study, the research draws on policy analysis and interviews to reveal governance obstacles in procurement and executive ICT engagement. KwaZulu-Natal shows progress due to strong leadership, while Eastern Cape and Free State face delays from compliance-driven cultures and bureaucracy. The study urges a balance between regulation and agility, recommending GITO empowerment through decentralized procurement and leadership development. It advances ICT governance theory by exposing multi-level implementation challenges.

A Python-based monitoring service for user metrics interacting with an open marketplace API has been developed. The architecture is implemented using Prometheus and Grafana and is focused on monitoring the performance of key stages of data processing: the number of requests, errors, response time, database write speed, and product characteristics. To assess system resilience, failure scenarios were simulated, including external API outages, database degradation, network delays, increased load, and memory leaks. The use of stream data processing in combination with SQLite ensures high performance and reliability.

This article is aimed to improve the robust synthesis of the discrete proportional-integraldifferential (PID) controller with first and second derivatives in the control law. The PID control tuning and synthesis methods are based, in many cases, on the experimental or online information without mathematical model. Also, the fuzzy PID control is in usage. The synthesis problem of PID control still actual, because of plant uncertainty and perturbations. Then, the enhanced PIDm control with m-order derivatives in the control law is offered in this article. The proposal is based on the modal control with high amplification in the PIDm control channel. That provide the domination of the controller parameters in the system when plant parameters are uncertain. However, in case of the discrete (digital) control, the linear system stability is bounded by the amplification values, so the additional analysis is required. The analysis is accomplished of the system with PID controllers in cases of m = 1 and m = 2. The expressions are established for PID controllers parametrization. The developed PID parametrization can be evaluate for m ≥ 2. The simulation is accomplished of the systems with proposed parametrization of PIDm controllers for m = 1 and m = 2 cases. The results of the simulation show the effectiveness of the proposed technique of synthesis in conditions of plant perturbations.

The article is devoted to computer modeling of the social and medical problem of determining the degree of obesity of body mass. An analysis was conducted, the significance of the solution to the problem was revealed. Using modern mathematical methods based on the theory of fuzzy sets and the method of fuzzy logical deduction, an approach is presented to determining the degree of obesity of body weight. Based on a short review, it was concluded that the body mass index, waist circumference size, degrees of abdominal obesity are recognized as direct factors in measuring the degree of obesity of body weight. Since all three factors are fuzzy parameters, the method of fuzzy logic inference is used to make a decision. The revealed universes of fuzzy variables of obesity of body weight, body mass index, waist circumference size, degrees of abdominal obesity. With the help of vague production rules, the answer options are provided for various situations. Fuzzy logical inference is implemented by the Mamdani method. The constructed model allows to calculate obesity rates for a large number of people in a short period of time objectively and accurately. Due to this feature, the developed approach can be used in population studies when defining specific categories of body weight as a health problem. The accumulated data of this kind are significant in life insurance. The results are related not only to adverse health problems, but also to social problems, such as determining the fatness of the population of different regions.

The paper presents a new method for automated no-reference quantitative assessment of the quality of digital retinal images for diabetic retinopathy screening. The proposed method does not require localization of anatomical structures and is based on the analysis of the central fragment of the image in the green spectral channel using the Weibull distribution scale parameter for integrating local quality estimates. A comparative analysis of 36 no-reference functions was carried out, two evaluation measures that showed the best results were selected. It was experimentally shown that using a central fragment 50–67% of the original image size allows increasing the accuracy of image quality assessment by 40 % compared to full-image analysis. Scaling this fragment to 512×512 pixels reduces the image analysis time by up to 20 times without losing accuracy. The effectiveness of the method was confirmed on three thousand images from various sources: Kaggle and DDR databases, Belarusian clinical data. The developed approach does not require reference data and can be integrated into mass screening systems of fundus images, reducing the workload of specialists and increasing the availability of diagnostics for patients with limited computing resources.

As a result of the conducted research, a method was developed to help people with the most common forms of dichromacy – deuteranopia and protanopia. For people with protanopia, the transformation of colors with a predominant red component is carried out, for deuteranopes – with a predominant green component. The features of the described method are finding colors indistinguishable for dichromats using a customizable colorimetric deviation and changing the coordinates b and L of indistinguishable colors in such a way that as a result of the transformations, the indistinguishable colors of one image differ from each other as much as possible, but at the same time retain their naturalness in the best possible way. Among the advantages of this method, it is worth noting the ability for each user to customize such personalized recoloring parameters as the colorimetric deviation of indistinguishable colors, the transformation coefficient of coordinate b, and the transformation coefficient of brightness L in accordance with their individual perception of visual information. Recoloring according to the developed method is acceptable for both more realistic images and for charts, signs, HTML-documents, and application interfaces. The correctness of the method was verified by examining a recolored image simulated by the Brettel et al. method for deuteranopic and protanopic vision with a normal trichromat, as a result of which areas of the image previously inaccessible to the dichromat's vision became distinguishable. The quality of recoloring was also assessed by the loss of color naturalness and global chromatic diversity. Thus, the loss of color naturalness for the test images had satisfactory values and varied within the range from 4.39 to 20.15 depending on the color component of the images. The global chromatic diversity of all test images was increased as a result of recoloring. The execution time of the method for both dichromacy cases indicates a satisfactory image processing speed and is 3 s for images of 170 000 pixels.

The purpose of the article is to study the using of technologies: machine learning (ML), neural networks (NN), Internet of Things (IoT) and blockchain (BC) to improve the effectiveness of education. Thestudyexaminesthe limitations of traditional educationandtheimpactof digitalization. Theadvantagesandsystems from the separate use of machine learning and neural networks are presented: predicting academic performance, analyzing student behavior, and verifying knowledge. The IoT architecture in education is considered, consisting of three levels: perception, network and applications. The process of integrating ML, NN, IoT, and BC technologies has been developed, including data collection using IoT devices, analytical data processing using ML and NN, and reliable data storage using blockchain. Based on this scheme, the structure of the integration system is proposed, consisting of modules for data collection, intelligent analysis and storage, confirmation, and data protection.