Text this: Constrained video face clustering using 1NN relations