Аннотация:The paper proposes a new method of generative augmentation of histological images based on the use of Generative Adversarial Networks (GANs). Based on a relatively small amount of labeled data, GANs are able to create new synthetic images with appropriate labeling, which can be used to train convolutional neural networks (CNNs) for histological image analysis. It is shown that this approach expands training dataset and, as a result, increases the accuracy of classification models on test data. It is shown that transfer learning and freezing the first layers of GANs improve their efficiency.