Description

Abstract

—  We propose a method to cluster-specific object detections of a video sequence, which we applied to face detections.

—  The proposed method is based on three main stages

—  First we use a face detector to localize faces in all frames of the video and extract various features of the detections.

—  These features are then used to compute a dissimilarity matrix based on appearance and space–time.

—  Finally, an optimization method involving a probabilistic model is employed to group all the detections according to the dissimilarity matrix.