Abstract:

Understanding human activity is a very challenging task, but a prerequisite for the autonomy of robots interacting with humans. Solutions that generalize must involve not only perception but also cognition and a grounding in the motor system. Our approach is to describe complex actions as events at multiple time scales. At the lowest level, signals are chunked into primitive symbolic events, and these are then combined into increasingly more complex events of longer and longer time spans. The approach will be demonstrated on our work of creating visually learning robots, and the talk will describe some of its novel components: an architecture that has cognitive and linguistic processes communicate with the vision and motor systems in a dialog fashion; vision processes that parse the objects and movements based on their attributes, spatial relations, and 3D geometry; the combination of tactile sensing with vision for better recognition; and approaches to cover long-term relations in observed activities.

Bio:

Cornelia Fermüller is a research scientist at the Institute for Advanced Computer Studies (UMIACS) at the University of Maryland at College Park.  She holds a Ph.D. from the Technical University of Vienna, Austria and an M.S. from the University of Technology, Graz, Austria, both in Applied Mathematics.  She co-founded the Autonomy Cognition and Robotics (ARC) Lab and co-leads the Perception and Robotics Group at UMD. Her research is in the areas of Computer Vision, Human Vision, and Robotics. She studies and develops biologically inspired Computer Vision solutions for systems that interact with their environment. In recent years, her work has focused on the interpretation of human activities, and on motion processing for fast active robots (such as drones) using as input bio-inspired event-based sensors.

http://users.umiacs.umd.edu/users/fer

 This talk will be recorded. Click Here for all of the recorded seminars for the 2019-2020 academic year.