Industrial Automatic Control Systems and Controllers Annotation << Back Approaches to the Development and Usage of 3D Models for Training Object Recognition Convolutional Neural Networks N.D. Shur, K.A. Barinov As part of the process of developing and preparing a technical solution using computer vision and convolutional neural networks (hereinafter referred to as CNN), one of the most time-consuming and resource-intensive processes is the preparation of training and testing datasets, used during the CNN training. Currently, several approaches are used to collect a dataset of images with the required volume. This article discusses the most innovative approach based on 3D-modeling and generation of synthetic images. In addition, the article presents the results of the experiment in the framework of the problem of detection of personal protective equipment from the employees of the enterprise by means of photo and video streaming. Keywords: artificial intelligence, computer vision, convolutional neural networks, neural network training, 3D-modeling, CAD. DOI: 10.25791/asu.8.2022.1381 Pp. 30-37.