Assoz. Prof. Martin Holler from the Department of Mathematics and Scientific Computing at the University of Graz will give a lecture on the topic “How to deal with the black box - uncertainty quantification for AI-based image reconstruction methods”.
In the past decade, machine learning has revolutionized the field of computational imaging, leading to unprecedented results in image quality across applications such as image super resolution or text-to-image generation. While in particular the latter has received a lot of media attention recently, the influence of machine learning also extends to classical image reconstruction problems in biomedical imaging, including MR, PET, or CT. Taking MR imaging as an example, in this talk Martin Holler will explain the main principles of successful machine-learning-based image reconstruction methods. He will then address the risks and shortcomings of these methods and discuss how they can be controlled using techniques from uncertainty quantification.
