Convolutional neural networks (CNNs) have been successfully used for image classification for over a decade. Previously, achieving good results required tens of thousands of images per class. Now, using few-shot and one-shot learning approaches, it is possible to obtain a model with decent performance even when only a few images per class are available. The seminar will cover general concepts of neural networks, convolutional neural networks, methods for data preparation and model training, approaches to training under few-shot learning conditions, and the results of studies related to plant disease classification.