|
Abstract:
|
The purpose of this study is to generate facial composites by interactive differential evolution (DE). Faces are modeled using the Active Appearance Model (AAM). DE is similar to an evolutionary algorithm and yields superior results in optimization problems. In this study, an interactive version of DE is implemented in such a way that candidate faces generated by the AAM is presented to the user and the user selects a subset of these faces based on their resemblance to the target face. Experimental results show that DE is successful in converging to a face rapidly and it is robust to fluctuations in the user's decisions.
|
