ПРОЕКТИРОВАНИЕ НЕЙРОННОЙ СЕТИ НА ПРИМЕРЕ РАСПОЗНАВАНИЯ РУКОПИСНЫХ ЦИФР

Нейронные сети

Abstract

This article is devoted to the implementation of an algorithm for recognizing handwritten digits using neural networks. Definition of a full-fledged neural network in Keras on a real example of handwritten digit recognition. Improving a full-fledged neural network in Keras by adding convolutional layers. An open database of MNIST handwritten digit images is considered as a test sample. The resulting model can be successfully used to solve problems of image classification and image recognition.