doc. dr. sc. Jelena Božek

This talk provides an overview of artificial intelligence (AI) applications in medical imaging, with a focus on the objectives of the ongoing FetalAI project funded by the Croatian Science Foundation (HRZZ). It will present how AI models are being developed for the early detection of fetal heart and brain anomalies. In addition to insights into the fetal and neonatal periods, the lecture will cover the application of algorithms for brain age estimation and the identification of patterns associated with mental disorders. Participants will learn how collaboration between engineers and clinicians enables more precise diagnostics, along with a discussion of the challenges of explainable artificial intelligence (XAI) in clinical practice.

Biography:

Jelena Bozek, PhD, is an Assistant Professor at the University of Zagreb Faculty of Electrical Engineering and Computing, with an extensive international research experience (University of Oxford, Radboud University). Her research focuses on advanced medical image analysis and the application of artificial intelligence in neuroscience and medicine. She has served as a principal investigator or collaborator on several scientific projects and has authored numerous publications. In addition to teaching at FER, she is a course coordinator for joint study programs in Applied Cognitive Science and Biomedical Engineering. She is a member of the IEEE, two bodies of the Croatian Academy of Sciences and Arts (the Committee for Neuroscience and the Scientific Council for Medicine and Technology), an Associate Editor for a scientific journal, and a member of the program committees of international conferences.

Email:

prof. dr. sc. Robert Cupec

Recent advances in foundation models have significantly changed how robots perceive and interact with the world. Instead of relying on task-specific pipelines and heavily labeled datasets, modern approaches leverage general-purpose models that can segment, recognize, and reason about visual scenes with minimal supervision.
This talk provides an accessible overview of how models such as Segment Anything Model, CLIP, DINO, and Depth Anything can be integrated into robotics systems to enable more flexible and generalizable perception. Practical capabilities, such as open-vocabulary object recognition, zero-shot segmentation, and monocular depth estimation, will be highlighted, and their integration for solving real robotic tasks like object manipulation and tracking, scene understanding, and navigation is demonstrated.
The session also discusses current limitations and open challenges, including robustness, real-time deployment, and the gap between perception and physical interaction. The goal is to provide students, researchers, and practitioners with an intuitive understanding of what these models enable today and where they may lead next in the development of more adaptable robotic systems.

Biography:

Robert Cupec graduated from the Faculty of Electrical Engineering, University of Zagreb, in 1995, where he also received his MSc degree in 1999. Upon graduation, he joined the Department of Control and Computer Engineering in Automation, University of Zagreb, where he was employed until 2000. From 2000 to 2004, he worked at the Institute of Automatic Control Engineering, Technische Universität München, where he received his PhD degree in 2005. Currently, he is employed as a Full Professor with tenure at the Faculty of Electrical Engineering, Computer Science and Information Technology Osijek, Josip Juraj Strossmayer University of Osijek, Croatia. His main research interest is in robot vision.

Email:

doc. dr. sc. Hrvoje Leventić

Email: