In this project we aim to recognize and to localize human actions in generic video such as feature films, documentaries, TV news and home recordings. Human actions are among the most informative events in the video. Recognition of human actions is therefore essential to the video search technology and its applications.

The following video demostrates our
results of detecting actions of class "drinking" in the movie "Coffee and Cigarettes" (2003) - Jim Jarmusch.

Further details

"Retrieving actions in movies",
I. Laptev and P. Pérez; in Proc. ICCV'07

On-line talk

Annotation for actions "drinking" and "smoking" used in ICCV07 paper
Video data for a set of drinking actions recorded at INRIA/Vista and used the ICCV07 paper
The summary of action annotation (as a mean of visualization only)