Unsupervised Classemes

TitleUnsupervised Classemes
Publication TypeConference Paper
Year of Publication2012
AuthorsCusano, C, Satta, R, Santini, S
Conference Name2012 ECCV Workshop on Information Fusion in Computer Vision for Concept Recognition (IFCVCR 2012)
Date Published13/10/2012
Conference LocationFirenze (Italy)
In this paper we present a new model of semantic features that, unlike previously presented methods, does not rely on the presence of a labeled training data base, as the creation of the feature extraction function is done in an unsupervised manner. We test these features on an unsupervised classification (clustering) task, and show that they outperform primitive (low-level) features, and that have performance comparable to that of supervised semantic features, which are much more expensive to determine relying on the presence of a labeled training set to train the feature extraction function.
Citation Key 1358
Cusano_IFCVCR2012.pdf259.36 KB