Towards Anticipation in Traffic Scene Understanding
Scene understanding e.g. in terms of semantic segmentations and object detections has made great advances in recent years. In order to facilitate safe autonomous or assisted driving in real-world scene, we have to go beyond assessing the current traffic scene and rather predict possible consequences and future states.
I will present our latest work in this direction that aims at “predicting the future”. In particular, we have been developing Deep Learning techniques that encode prior observations and decode them in a recursive fashion and thereby extrapolate them into the future.
Mario Fritz is senior researcher at the Max Planck Institute for Informatics. He is heading a group on Scalable Learning and Perception. His research interest are centred around computer vision and machine learning but extend to natural language processing, robotics, graphics, HCI, privacy and more general challenges in AI. He did his postdoc at the International Computer Science Institute as well as UC Berkeley on a Feodor Lynen Research Fellowship of the Alexander von Humboldt Foundation. He received his PhD from TU Darmstadt and graduated from the University of Erlangen-Nuremberg.
ACM Chapters Computer Science in Cars Symposium 2017