Evaluating the performance of Face-Aging Algorithms

Abstract

An important aspect related to the development of faceaging algorithms is the evaluation of the ability of such algorithms to produce accurate age-progressed faces. In most studies reported in the literature, the performance of face-aging systems is established based either on the judgment of human observers or by using machine-based evaluation methods. In this paper we perform an experimental evaluation that aims to assess the applicability of human-based against typical machinebased performance evaluation methods. The results of our experiments indicate that machines can be more accurate in determining the performance of face-aging algorithms. Our work aims towards the development of a complete evaluation framework for age progression methodologies.