Speaker: Ing. Maroš Kollár
Abstract
This presentation provides an overview of the group’s research activities, with an emphasis on the use of generative neural networks, such as Generative Adversarial Networks (GANs) and Diffusion Models, in the medical domain. The focus is placed on the generation of synthetic data to augment medical datasets, aiming to improve the performance and robustness of downstream tasks like classification and segmentation. We also explore the potential of these generative approaches to address challenges related to data scarcity, variability, and annotation costs in medical imaging.
Bio
Ing. Maroš Kollár is a PhD student whose research focuses primarily on the use of generative neural networks, such as Generative Adversarial Networks (GANs) and diffusion models, in the fields of computer graphics and histology. His conference publications, titled “Multiclass Texture Synthesis Using Generative Adversarial Networks” and “Semantically Controlled Texture Synthesis by Diffusion Model,” address the topic of image generation using generative neural networks.