The Use of Convolutional Neural Networks in Multiclass Classification of Digital Pictures

The article shows a problem of digital picture classification with the use of convolutional neural networks in a multiclass classification problem. It describes basic problems related to the classification of data sets with a very large number of classes, and basic picture preprocessing operations. It presents the convolutional neural network model in use, its training procedure and the results of the experiment on a set containing a large number of classes – Stanford Dogs.