Anfonwch hwn fel neges destun: Constrained video face clustering using 1NN relations