EU flag and ERC logo
Prof. Daniel Cremers will use video data to generate the most detailed possible model of the observed world. The simulation of elastically deformable shapes is an important tool for this purpose. The motion phases of the cartoon armadillo were created using a deformable shape of this kind. Image: Eisenberger, Lähner, Cremers / TUM
Prof. Daniel Cremers will use video data to generate the most detailed possible model of the observed world. The simulation of elastically deformable shapes is an important tool for this purpose. The motion phases of the cartoon armadillo were created using a deformable shape of this kind.
Image: Eisenberger, Lähner, Cremers / TUM

, Medizinische Bildverarbeitung & KI, News

Physics modeling using video data

ERC Advanced Grant awarded to MSB-P Daniel Cremers

Daniel Cremers, Professor of Computer Vision and Artificial Intelligence at the Technical University of Munich and Prinicipal Investigator at the Munich School of BioEngineering, is to receive funding in the form of an Advanced Grant from the European Research Council (ERC). In his ERC-funded project, Prof. Cremers wants to develop new algorithms to physically simulate deformable objects. 

ERC Advanced Grants, which are set aside for established, leading scientists with a track record of significant research achievements over the past 10 years, come with up to 2.5 million euros in funding. In this round, two researchers at the technical University of Munich have been awarded an advanced grant.
 

Physics modeling using video data

It is already possible to gain large amounts of information on the position of objects in space using video data. With his SIMULACRON project, Prof. Daniel Cremers now aims to use video information to determine physical properties such as acceleration, mass and elasticity. Humans can assess many of these properties very quickly. For example, it takes us just a fraction of a second for a person to predict the path of a tennis ball.
To enable computers to do the same, Prof. Cremers wants to develop new algorithms to physically simulate deformable objects. The simulation parameters will be determined directly from video images, also using new machine learning techniques. SIMULACRON supplies the machine with a more complete understanding of the observed world based on a small quantity of observed data. This technology could be used in robots and autonomous vehicles, for example.