Ressource pédagogique : Action recognition from video: some recent results

cours / présentation - Date de création : 01-12-2011
Auteur(s) : Cordelia SCHMID
Partagez !

Présentation de: Action recognition from video: some recent results

Informations pratiques sur cette ressource

Langue du document : Anglais
Type pédagogique : cours / présentation
Niveau : master, doctorat
Durée d'exécution : 44 minutes 9 secondes
Contenu : image en mouvement
Document : video/mp4
Taille : 96.59 Mo
Droits d'auteur : libre de droits, gratuit
Droits réservés à l'éditeur et aux auteurs.

Description de la ressource pédagogique

Description (résumé)

While recognition in still images has received a lot of attention over the past years, recognition in videos is just emerging. In this talk I will present some recent results. Bags of features have demonstrated good performance for action recognition in videos. We briefly review the underlying principles and introduce trajectory-based video features, which have shown to outperform the state of the art. These features are obtained by dense point sampling in each frame and tracking them based on displacement information from a dense optical flow field. Trajectory descriptors are obtained from motion boundary histograms, which are robust to camera motion. We then show how to integrate temporal structure into a bag-of-features model based on so-called actom sequences. We localize actions based on sequences of atomic actions, i.e., represent the temporal structure by sequences of histograms of actom-anchored visual features. This representation is flexible, sparse and discriminative. The resulting model is shown to significantly improve performance over existing methods for temporal action localization. Finally, we show how to move towards more structured representations by explicitly modeling human-object interactions. We learn how to represent human actions as interactions between persons and objects. We localize in space and track over time both the object and the person, and represent an action as the trajectory of the object with respect to the person position, i.e., our human-object interaction features capture the relative trajectory of the object with respect to the human. This is shown to improve over existing methods for action localization.

"Domaine(s)" et indice(s) Dewey

  • Vision par ordinateur (006.37)

Thème(s)

Intervenants, édition et diffusion

Intervenants

Fournisseur(s) de contenus : INRIA (Institut national de recherche en informatique et automatique), UNS, CNRS - Centre National de la Recherche Scientifique, VSP - Vidéo Sud Production

Éditeur(s)

Diffusion

Document(s) annexe(s) - Action recognition from video: some recent results

Partagez !

AUTEUR(S)

  • Cordelia SCHMID

ÉDITION

Région PACA

INRIA (Institut national de recherche en informatique et automatique)

EN SAVOIR PLUS

  • Identifiant de la fiche
    7945
  • Identifiant
    oai:canal-u.fr:7945
  • Schéma de la métadonnée
  • Entrepôt d'origine
    Canal-U
  • Date de publication
    01-12-2011