Application of convolutional neural networks for problems of fruits classification in the image

Abstract

The method for solving a specialized problem of pattern recognition - the problem of classifying fruits on images by using a multilayered convolutional neural network (CNN) has been proposed in the article. General information about the mechanism of CNN work has been considered. The description of the neural network ResNet chosen for the task solution is given. The presented method for creating a software system on Python that can perform the task for classifying 30 classes of images with fruits allows performing the task of subsequent marking of images containing several objects. The conclusions about the applicability of the ResNet neural network for the classification of the required data set are presented. The accuracy metrics of the selected architecture are shown.