Group detection represents an emerging Computer Vision research topic, motivated by the increasing interest in the modelling of the social behaviour of people. This paper presents an unsupervised method for group detection which is based on an on-line inference process over Dirichlet Process Mixture Models. Formally, groups are modelled as components of an infinite mixture and individuals are seen as observations generated from them. A sequential variational framework allows to perform the inference in real-time, while social psychology constraints of proxemics ensure the production of proper group hypotheses, consistent with the human perception. The results obtained on different compare favorably with state-of-the-art approaches, setting the best performance in some of them.
Online Bayesian Non-parametrics for Social Group Detection
BAZZANI, Loris;CRISTANI, Marco;MURINO, Vittorio
2012-01-01
Abstract
Group detection represents an emerging Computer Vision research topic, motivated by the increasing interest in the modelling of the social behaviour of people. This paper presents an unsupervised method for group detection which is based on an on-line inference process over Dirichlet Process Mixture Models. Formally, groups are modelled as components of an infinite mixture and individuals are seen as observations generated from them. A sequential variational framework allows to perform the inference in real-time, while social psychology constraints of proxemics ensure the production of proper group hypotheses, consistent with the human perception. The results obtained on different compare favorably with state-of-the-art approaches, setting the best performance in some of them.
I documenti in IRIS sono protetti da copyright e tutti i diritti sono riservati, salvo diversa indicazione.
