What happens to AI when we stop thinking on flat ground?

Prof. Dr. Daniel Cremers—ELIZA Academic Fellow and Chair of Computer Vision and Artificial Intelligence at the Technical University of Munich—has built a career at the intersection of rigorous mathematics, computer vision, and real‑world impact. From early work on shape analysis and optical flow to award‑winning contributions in medical imaging and robotics, his research consistently asks how geometry can deepen what machines can perceive and infer.

In his recent appearance in the AI for Good From Molecules to Models series, Cremers brings this perspective to the webinar “Convolutional networks beyond the Euclidean space.” Rather than treating convolutional networks as a solved chapter of deep learning, he invites the audience to consider what happens when data lives on graphs, manifolds, or other non‑Euclidean structures—and why this matters for applications like medical diagnosis and molecular science.​​​