Thomas Brox

We propose a world model for autonomous driving capable of long-term rollouts in challenging scenarios, achieving state-of-the-art-results. The model is trained for next frame prediction on video data.