A seminar titled " Activity Recognition for Real-world Scenarios " is held in KAIST.
Title : Activity Recognition for Real-world Scenarios
Date : September 24 (Monday), 2012 14:00
Location : Wooribyul Seminar Room(#2201), Department of Electrical Engineering
Speaker : Dr. Michael S. Ryoo / NASA's Jet
Propulsion Laboratory
Abstract: Human activity recognition, an automated detection of ongoing events from video data, is an important research area obtaining an increasing amount of public interest. Its applications include surveillance systems, content-based web-video search, and a variety of systems that involve interactions between humans and machines such as robots and autonomous vehicles. In this talk, we particularly focus on an important activity recognition problem crucial for real-world systems: human activity prediction. The objective of activity 'prediction' is to identify an ongoing activity in its early stage (i.e., before it is fully progressed) from streaming videos, which is essential to prevent dangerous activities and accidents from happening. We introduce the prediction problem, present the prediction methodologies using spatio-temporal features, and discuss their applications.