Unsupervised Video Understanding by Reconciliation of Posture Similarities