Real Virtual Humans (RVHu)

• Applicant: Professor Dr. Gerard Pons-Moll
• Subject Area: Image and Language Processing, Computer Graphics and Visualisation, Human Computer Interaction, Ubiquitous and Wearable Computing
• Term: since 2018
• Project identifier: Deutsche Forschungsgemeinschaft (DFG) - Project number 409792180

Virtual humans are at the centre of diverse application areas such as medicine and psychology, virtual and augmented reality, and special effects in movies. For man-machine interaction it is crucial to develop models of humans that look and move indistinguishably from real humans. To enable natural communication, machines need to appear human-like to us. Machines also need detailed internal representations of real people in order to perceive (from sensory data) all subtleties that make us real. To achieve realism, all details matter: the facial expressions, the geometry of the body and its movements, the soft-tissue motion, the appearance and dynamics of clothing or the light reflecting on our body. All these are components that need to be modelled and perceived to the highest precision. Our hypothesis is that easier to control and more realistic models of humans and clothing can be learned by capturing real people using 3D/4D scans, videos and inertial sensors. The aim of this project is to build compact and rich representations of people and develop inference techniques to extract such representations from visual data. The two key research questions we want to address are:1) How do we efficiently digitize humans without losing the detail that make us real? 2) Given a compact digitization, how can we train machines to perceive such rich representations from visual data?

• DFG Programme: Independent Junior Research Groups