Learning temporal, relational, force-dynamic event definitions from video
Alan Fern, Jeffrey Mark Siskind, Robert L. Givan · 2002
We present and evaluate a novel implemented approach for learning to recognize events in video. First, we introduce a sublanguage of event logic, called k-AMA, that is suffi-ciently expressive to represent visual events yet sufficiently restrictive to support learning. Second, we develop a specific-to-general learning algorithm for learning event definitions in k-AMA. Finally, we apply this algorithm to the task of learn-ing event definitions from video and show that it yields defi-nitions that are competitive with hand-coded ones.