Deep Generative Models

Generative Adversarial Networks (GANs) and similar methods (e.g. Diffusion Models, Variational Auto Encoders) have carved open many exciting ways to tackle well known and challenging medical image analysis problems such as medical image denoising, reconstruction, segmentation, data simulation, detection or classification. Furthermore, their ability to synthesize images at unprecedented levels of realism also gives hope that the chronic scarcity of labeled data in the medical field can be resolved with the help of these generative models. This class starts with basic statistical concepts of image generation. We proceed with several lectures about contemporary deep learning approaches for image generation, such as Diffusion Models, Neural Cellular Automata, VAEs and GANs. As this is an integrated lecture, students can participate in a voluntary paper presentation and/or programming project to gain practical experience and an exam bonus. The lecture takes place in every summer term.