Understanding and Anticipating Activities

Last week, Professor Dr. Juergen Gall led a talk about the aspects of video understanding. He began speaking about holistic video understanding, where a video can be categorized by the activity as well as other tags that can describe scenes or attributes of video. He goes on to describe a multi-stage temporal convolutional network, which segments long videos that have to be processed in applications, such as surveillance or robots. However, just looking at the past is not enough anymore. Dr. Gall shows how robots need to anticipate the future activities of people, and discusses how this can actually be done. Throughout the talk, Dr. Gall talks in detail about these topics and presents the research he has been a part of throughout his career.

Dr. Gall’s research and presentation is about how robots and machines can process and understand video, and how this can relate to understanding human action. Ending his talk with the discussion of what is needed for close human-robot collaboration, Dr. Gall shows how this collaboration is possible. For machines, including surveillance or robots, to make decisions or communicate with people, they must be able to anticipate future human action and behavior. As machines become more able to do this, they can become smarter and more able to reach a capacity that is closer to humans. Anticipating and understanding activities and human action is an incredibly important aspect of AI research and development. 

Professor Dr. Juergen Gall is a professor at the University of Bonn, working in the Department of Information Systems and Artificial Intelligence. He received his PHD in Computer Science at the Universität des Saarlandes and the Max-Planck-Institut. He previously worked at the Max Planck Institut as well as the Computer Vision Laboratory, ETH Zurich. He is a computer vision researcher, known for activity recognition, human pose recognition, and object detection and segmentation. He has over 10,000 citations and an extensive network of researchers and work.