AI ON RAILS: BUILDING A SELF-SUPERVISED FOUNDATION MODEL FOR SWEDEN’S RAILWAYS

​Physical AI on rails! 🚂

​Trafikverket captures hundreds of terabytes of high-resolution imagery every year to monitor Sweden’s railway network. But there’s a problem: at that scale, manually labeling the data simply isn’t feasible.

In this talk, we’ll share how we’re turning this vast archive of unlabeled imagery into a useful AI resource using self-supervised learning. We’ll explore how a foundation model can learn meaningful visual representations directly from railway data, and how those representations can then be adapted to real-world tasks with only a small amount of labeled data.

Using examples such as detecting rail defects and identifying infrastructure assets, we’ll show what it takes to move self-supervised learning from research into a large-scale industrial setting—and what we’ve learned along the way.

​Agenda:

​14:00 Doors open, coffee and registration

​14:15 - 14:45 Presentation

​14:45 - 15:15 Questions and networking

​Presentation by eghed - experts in Data Science and AI

​www.eghed.se

​

​This event is part of Nordic Tech Week. The event is independently organised by the host. If you have any questions about Nordic Tech Week, please contact events@nordictechweek.org. Questions about this event should be directed to the event host.