A multiple labeling-based optimum-path forest for video content classification

Abstract

Multiple-labeling classification approaches attempt to handle applications that associate more than one label to a given sample. Since we have an increasing number of systems that are guided by such assumption, in this paper we have presented a multiple-labeling approach for the Optimum-Path Forest (OPF) classifier based on the problem transformation method. In order to validate our proposal, a multi-labeled video classification dataset has been used to compare OPF against three other classifiers and another variant of the OPF classifier based on a k-neighborhood. The results have shown the validity of the OPF-based classifiers for multi-labeling classification problems. © 2013 IEEE.