Actualités
  Période
semaine
ou mois
aucune période


  Thèmes:
Thèses
Colloques
Autres
Groupes
Tous les thèmes

   
  Liens:
Voir les thèses

Voir les colloques

Voir les autres événements

Voir la page des groupes

Accéder à l'intranet

Intitulé:   
Category level object segmentation by combining Bag-of-Words models with Dirichlet Processes and Random Fields
du groupe Image et Son

Date 2010-04-08  09:15-10:00
TitreCategory level object segmentation by combining Bag-of-Words models with Dirichlet Processes and Random Fields  
RésuméBag-of-words models successfully predict the presence of an object within an image; however, they can not accurately locate object boundaries. Random Fields take into account the spatial layout of images and provide local spatial regularization. Yet, as they use local coupling between image labels, they fail to capture larger scale structures needed for object recognition. In this presentation I will show how these components (Bag-of-words and MRF) can be combined with a Dirichlet process mixture in a single model. Images are modeled as a composition of regions, each representing a single object instance. Gibbs sampling is used for parameter estimations and object segmentation. This model successfully segments object category instances, despite cluttered backgrounds and large variations in appearance and viewpoints. A certain number of points will also be discussed: first, different methods to build visual vocabularies are compared. Second, we will see how to combine strong labeling (segmented images) with weak labeling (images annotated with bounding boxes), in order to limit the labeling effort needed to learn the model. Finally, the effect of different initializations is studied.  
LieuAmphi du LaBRI 
OrateurDiane Larlus 



Aucun document lié à cet événement.

Retour
Retour à l'index