Jet Propulsion Laboratory, NASA, USA
First-Person Activity Recognition: What Are They Doing and What Will They Do to Me?
Abstract: A massive amount of first-person videos is becoming available. Thanks to the recent development of small wearable cameras, humans (and robots) are now gaining an ability to constantly obtain first-person videos equivalent to ‘what they are seeing’. The paradigm is changing from ‘cameras you carry and turn on to record other people’s movements’ to ‘cameras you wear everyday to constantly record/analyze your own events from a first-person perspective’. In this talk, we introduce and discuss the concept of ‘first-person activity recognition’, the problem of recognizing human activities from such egocentric videos. We present video features and algorithms necessary for first-person recognition, while particularly focusing on human interactions such as ‘a person punching the observer (e.g., a robot)’ , ‘a person shaking hands with the observer’, and ‘a person throwing objects at the observer’. Approaches for prediction (i.e., early recognition) of ongoing activities from streaming first-person videos will also be described, and its future directions and applications will be discussed.
About the speaker: Michael S. Ryoo is a research staff at NASA’s Jet Propulsion Laboratory. His research interest includes semantic understanding of videos, first-person computer vision (first-person activity recognition particularly), and intelligent interactions/collaborations between humans and wearbles/robots. Dr. Ryoo received the B.S. degree in computer science from Korea Advanced Institute of Science and Technology (KAIST) in 2004, and the M.S. and Ph.D. degrees in computer engineering from the University of Texas at Austin in 2006 and 2008, respectively. He has been providing tutorials on human activity recognition at major Computer Vision conferences including CVPR 2011, AVSS 2012, and CVPR 2014, has authored more than 30 conference/journal papers on human activity recognition, and is the corresponding author of the activity recognition survey paper published by ACM Computing Surveys on 2011. He also was the lead organizer of the first ICPR contest on human activity recognition (SDHA 2010) and organized the 3rd workshop on Egocentric (First-person) Vision at CVPR 2014.