Сервис-ориентированный подход к разработке систем на базе свёрточных нейронных сетей

Abstract

Purpose: to substantiate the relevance of creating an automated system for generating services based on convolutional neural networks, using models and methods of service-oriented architecture. Described an approach to creating such a system based on deep learning. Materials and methods: the article describes the architecture and applications of an automated system for generating services in the concept of service-oriented architecture and training convolutional neural networks based on a genetic algorithm. Results: testing (validation) of the presented system was carried out on the example of solving the problem of recognizing reindeer from aerial photography. Conclusions: the advantages and disadvantages, implementation features, areas of application of the presented system are shown.