The Power and Pitfalls of Vision Transformers

Ved Line Petersen og Jonas Gjerris, Capacit

Vision Transformers have redefined computer vision, outperforming traditional models across domains. But what makes them so powerful, and why is self-supervised learning key to their success? We will explore how Vision Transformers “see” the world, where they excel and where they fall short. 

Through real-world examples and cutting-edge models, gain insights into the trade-offs between open-source solutions and custom-trained models – and discover how to leverage their full potential in your own applications.

Bio: Jonas and Line are AI Tech Lead and AI Engineering Manager at Capacit, where they specialise in developing custom AI solutions that help businesses leverage the latest advancements in AI. Beyond their industry work, they collaborate with the AI Pioneer Center on research in Vision Transformers and self-supervised learning, with a particular focus on ultrasound imaging.

Optagelsen stammer fra dette års Driving AI afholdt i København den 20. maj 2025, hvor 250 AI-professionelle og entusiaster mødtes for at høre om nye tendenser og teknologier inden for AI.