The demand for online education is gradually growing. Most universities and other institutions are faced with the fact that it is almost impossible to track how honestly test takers take exams remotely. In online formats, there are many simple opportunities that allow for cheating and using the use of outside help. Online proctoring based on artificial intelligence technologies in distance education is an effective technological solution to prevent academic dishonesty. This article explores the development and implementation of an online control proctoring system using artificial intelligence technology for conducting online exams. The article discusses the proctoring systems used in Kazakhstan, compares the functional features of the selected proctoring systems, and describes the architecture of Proctor SU. A prototype of the Proctor SU proctoring system has been developed. As a pilot program, the authors used this system during an online university exam and examined the results of the test. According to the author’s examination, students have a positive attitude towards the use of Proctor SU online proctoring. The proposed proctor system includes features of face detection, face tracking, audio capture, and the active capture of system windows. Models CNN, R-CNN, and YOLOv3 were used in the development process. The YOLOv3 model processed images in real time at 45 frames per second, and CNN and R-CNN processed images in real time at 30 and 38 frames per second. The YOLOv3 model showed better results in terms of real-time face recognition. Therefore, the YOLOv3 model was implemented into the Proctor SU proctoring system.

The article analyzes the possibility and rationality of using proctoring technology in remote monitoring of the progress of university students as a tool for identifying a student. Proctoring technology includes face recognition technology. Face recognition belongs to the field of artificial intelligence and biometric recognition. It is a very successful application of image analysis and understanding. To implement the task of determining a person’s face in a video stream, the Python programming language was used with the OpenCV code. Mathematical models of face recognition are also described. These mathematical models are processed during data generation, face analysis and image classification. We considered methods that allow the processes of data generation, image analysis and image classification. We have presented algorithms for solving computer vision problems. We placed 400 photographs of 40 students on the base. The photographs were taken at different angles and used different lighting conditions; there were also interferences such as the presence of a beard, mustache, glasses, hats, etc. When analyzing certain cases of errors, it can be concluded that accuracy decreases primarily due to images with noise and poor lighting quality.

This research investigates the implementation of Kolmogorov–Arnold networks (KANs) for image processing in resource-constrained IoTs devices. KANs represent a novel neural network architecture that offers significant advantages over traditional deep learning approaches, particularly in applications where computational resources are limited. Our study demonstrates the efficiency of KAN-based solutions for image analysis tasks in IoTs environments, providing comparative performance metrics against conventional convolutional neural networks. The experimental results indicate substantial improvements in processing speed and memory utilization while maintaining competitive accuracy. This work contributes to the advancement of AI-driven IoTs applications by proposing optimized KAN-based implementations suitable for edge computing scenarios. The findings have important implications for IoTs deployment in smart infrastructure, environmental monitoring, and industrial automation where efficient image processing is critical.

Dealing with agriculture, it is valuable to know an amount of moisture in a soil and to know how to forecast the stored soil moisture within particular period. Forecasting the stored soil moisture works for planning an extent and structure of crop production areas and adjustment of plant-growing programs. Having known about an amount of moisture in one-meter soil and the depth of precipitation in a vegetation season shall help farmers to determine a seeding time, type of fertilizers depending on soil quality and to work out an irrigation schedule as well. In this regard, over the last few years some vigorous activities applied to machine training methods of the weather forecast have been launched in the world. The goal of present research is to develop an artificial neuron network which shall afford an opportunity to figure out a stored soil moisture prior to outgoing to winter in a short-term. North Kazakhstan Region agrometeorological measuring stations for the period from 2012 to 2022 were used in the course of the neuron network training. The Levenberg-Marquardt algorithm aimed at non-linear regression models optimization was chosen for network training. The algorithm includes sequential approximation of initial parameter values to a local optimum. The mean squared error (MSE) function and the correlation coefficient ensure accuracy and precision of forecasts. As a result, 7 neural networks under MATLAB environment using the Levenberg-Marquardt algorithm, with different input and output data, and with different number of learning iterations came to realization. Following analysis of the results, the choice was fallen on the ANN9 best network offering minimum error function and actual data maximum correlation. The neural network obtained fits for use to make efficient decisions in the North Kazakhstan region agricultural sector in the short term.

This study addresses the critical challenge of optimizing drone flight parameters for enhanced object detection in video streams. While most research focuses on improving detection algorithms, the relationship between flight parameters and detection performance remains poorly understood. We present a novel approach using Kolmogorov–Arnold Networks (KANs) to model complex, non-linear relationships between altitude, pitch angle, speed, and object detection performance. Our main contributions include the following: (1) the systematic analysis of flight parameters’ effects on detection performance using the AU-AIR dataset, (2) development of a KAN-based mathematical model achieving R² = 0.99, (3) identification of optimal flight parameters through multi-start optimization, and (4) creation of a flexible implementation framework adaptable to different UAV platforms. Sensitivity analysis confirms the solution’s robustness with only 7.3% performance degradation under ±10% parameter variations. This research bridges flight operations and detection algorithms, offering practical guidelines that enhance the detection capability by optimizing image acquisition rather than modifying detection algorithms.

Face anti-spoofing is crucial for protecting biometric authentication systems. Presentation attacks using 3D masks and high-resolution printed images present detection challenges for existing methods. In this paper, we introduce a family of specialized CNN architectures, AttackNet, designed for robust face anti-spoofing with optimized residual connections and activation functions. The study includes the development of four architectures: baseline LivenessNet, AttackNetV1 with concatenation-based skip connections, AttackNetV2.1 with optimized activation functions, and AttackNetV2.2 with efficient addition-based residual learning. Our analysis demonstrates that element-wise addition in skip connections reduces parameters from 8.4 M to 4.2 M while maintaining performance. A comprehensive evaluation was conducted on four benchmark datasets: MSSpoof, 3DMAD, CSMAD, and Replay-Attack. Results show high accuracy (approaching 100%) on the 3DMAD, CSMAD, and Replay-Attack datasets. On the more challenging MSSpoof dataset, AttackNetV1 achieved 99.6% accuracy with an HTER of 0.004, outperforming the baseline LivenessNet (94.35% accuracy, 0.056 HTER). Comparative analysis with state-of-the-art methods confirms the superiority of the proposed approach. AttackNetV2.2 demonstrates an optimal balance between accuracy and computational efficiency, requiring 16.1 MB of memory compared to 32.1 MB for other AttackNet variants. Training time analysis shows twice the speed for AttackNetV2.2 compared to AttackNetV1. Architectural ablation studies highlight the crucial role of residual connections, batch normalization, and suitable dropout rates. Statistical significance testing verifies the reliability of the results (p-value < 0.001). The proposed architectures show excellent generalization ability and practical usefulness for real-world deployment in mobile and embedded systems.