|
Evènement pour le groupe Séminaire du LaBRI
Date | 2010-11-04 14:00-15:00 |
Titre | Sém LaBRI. Recognition and Tracking of Human Actions in Image Sequences |
Résumé | Recent years have witnessed an unprecedented interest in the
visual analysis of human motion that is stimulated by
applications that include computer games, animated films and
entertainment, human computer interaction, surveillance/security
and multimedia retrieval. In this talk we will present an object
tracking scheme in which the predictions of multiple observations
are weighted by a probabilistic classifier that determines our
expetation that the corresponding prediction is relevant/useful
for the estimation of the state (position/scale/rotation) of the
target. This is particularly useful in the case of large motions
and occlusion, where many observations fall in areas that are far
from the target. The proposed scheme derives both generative and
discriminative tracking methods as special cases. We show the
benefits in accuracy in presence of persistent occlusions and
motion large in magnitude and briefly discuss the related
applications. Time permitting we will present our recent work in
discriminative action recognition and localisation, that is based
on a part-based implicit shape model in which local parts vote
for the location and scale of an action. |
Lieu | Amphi LaBRI |
Orateur | Ioannis Patras |
Url | Queen Mary University, UK |
Aucun document lié à cet événement. RetourRetour à l'index
| |