Exemplar Extraction using Spatio-Temporal Hierarchical Agglomerative Clustering for Face Recognition in Video