Maroš Kollár
Supervisor(s): Dr. Lukáš Hudec
Slovak Technical University
Abstract: Generative adversarial networks as a tool for generating content is currently one of the most popular methods for content synthesis. Despite its popularity, current solutions suffer from a drawback of shortage of generality. It means that trained models are usually able to synthesize only one specific kind of output. The usual texture synthesis approach for generating N different texture species requires training of N models with changing training data. In our work, we present a synthesis architecture model constraining and forcing the optimization for generating multiple texture types. We focus on the synthesis of realistic natural non-stationary textures. The solution allows users to control the class of texture to synthesize. Thanks to the controllable selections from feature space of synthesized texture, we also explore the possibilities of transitions between classes of trained textures for potential better usage in applications where texture synthesis is required.
Keywords: Computer Vision, Image Processing
Full text: Year: 2022
