Style-transfer v oblasti obrazových dat

Abstract

The main goal of the bachelor thesis was to create an application that allows the user to try styling images using style transfer methods. This application consists of two main parts, namely the server-side part and the user interface part. In the server-side part, there are the models that are used to create the actual styling and the API that we use to communicate with the user interface. The web interface allows the user to create stylized images using different models and also set their parameters. The thesis includes a summary of the history and development of style transfer methods for the image data domain. Furthermore, the thesis describes the use of convolutional neural networks for this problem, their functionality and use in the resulting application. The thesis also contains testing of selected model with the use of the developed application, where we can see sample stylizations for each of the selected models.