Laura Sevilla

I'm a Lecturer (Assistant Professor) at the School of Informatics in the University of Edinburgh as of January 2019. 

My research focuses on motion in the visual world: what motion tells us about the properties, the state and the structure of the world, and how to estimate it. This broadly includes motion estimation, temporal representations, video understanding, high-level perceptual tasks that involve motion and the machine learning problems that all these problems pose.

Before I joined Edinburgh, I was a postdoc at Facebook Research, and before I was at the Max Planck Institute for Perceiving Systems. 


  • If you're looking for a benchmark to test how good your model is at capturing temporal information in videos, try our new Temporal Dataset, featured in the Facebook AI blog
  • Our paper describing FASTER was just accepted to AAAI '19! Congrats Linchao. FASTER maintains the accuracy of state-of-the-art in video classification while reducing an order of magnitude the computation time. 
  • I'm co-leading with Yannis Kalantidis the very exciting CVPR Worksop on "Computer Vision for Global Challenges" , where we explore novel vision problems to humanitarian, development, and other global domains. Check our video below
  • We're presenting the tiny DMC-Net for Action Recognition at CVPR '19. Everything you wanted in an action recognition network, for a fraction of the cost. 
  • Kiyoon Kim and Shreyank Gowda to join the lab and start their PhD in September. 

I'm always looking for very motivated, independent and talented PhD students. 

CV4GC Impressions








lsevilla at exseed dot ed dot ac dot uk

Informatics Forum, 10 Crichton St, Edinburgh EH8 9AB