Keynote Speeches

Keynote Speech I

Pixels and Predictions: Perceiving What Is and What Could Be

João Carreira
Google DeepMind
London, UK

Vision has undergone tremendous progress recently with many classic problems solved. This talk will cover recent progress from our team at Google DeepMind around two areas — a) advanced visual capabilities including 4D reconstruction (3D + motion), multitask vision models and city-scale perception and b) going beyond perception and modelling the future, which is important for acting and robotics.

Biosketch

João Carreira is a Research Scientist at Google DeepMind, working on computer vision. He completed his PhD at the University of Bonn in 2012 working on object segmentation, advised by Cristian Sminchisescu, then was a PostDoc in sunny Berkeley with Jitendra Malik, working on 3D and human pose estimation. His focus at DeepMind has been on learning from video, and was one of the core creators of the Kinetics dataset and more recently the Perception Test. On the modelling side he was an author of the I3D and Perceiver models and is interested in developmental-inspired approaches that succeed at learning from continuous video streams, like people and animals do.