Detect Regularities & Anomalies in Events?
Note only can we detect motions of individual people, we can also automatically classify actions in a site.
If we go back to our outdoor setting, here are a set of trajectories taken over a period of a day. Each image represents the view for a one hour interval, and the colored patterns are the tracks of the detected moving objects, with color encoding direction, and brightness encoding speed. Our goal is to create computer systems that can automatically classify these trajectories into common events, and to record information about the frequency and execution of those events.