Video Understanding for Activity Recognition

The task of Video Understanding to recognize human activities has been studied for several decades, but it remains an open-problem due to the
complexity of video events (e.g., high-intra class variance complemented by low inter-class variance). A complex activity may be seen as a composition of two or more sub-parts, where a sub-part could be atomic (e.g., actions and objects, namely a concept) or another activity. In this context, we will discuss about activity recognition approaches which aim at analysing human behaviors in various environment, in public places for video-surveillance or at home for home-care to monitor older adults.

François Brémond Research Director INRIA