Instruments and Systems: Monitoring, Control, and Diagnostics Annotation << Back An Automated Image Augmentation System for Training Computer Vision Models V.A. SOBOLEVSKII, I.S. GROMOV In post-industrial economies, particularly in large industrial sectors, the use of computer vision systems is becoming an important application tool that helps solve many problems. The operation of such systems, as a rule, not only often meets the requirements of the subjects of the industrial economy using them but also has several important conditions for proper functioning and successful solutions of tasks. The key problem in creating computer vision systems is to collect the necessary amount of training data. In the industrial sector, this problem is particularly acute, as it requires monitoring of industrial facilities and processes that are not found or used in everyday life. In this regard, when developing computer vision systems in industry, it is often impossible to use open arrays of training data, since either no one collects them, or they are protected by a trade secret. This article proposes a solution to the problem of image scarcity for the creation of industrial computer vision systems using an automated image augmentation system. The system uses a weighted average estimate of the geometric, color, noise, and compositional properties of the original images, based on which it generates new ones in the amount specified by the user. The developed system integrates the most effective algorithms for image evaluation and augmentation, and also has functions for working with labeled images, allowing you to expand the available variety of photography for computer vision tasks. Keywords: computer vision, augmentation, deep learning, machine learning, monotony. DOI: 10.25791/pribor.11.2025.1629 Pp. 26-35.