Interactive Nature-Inspired Heuristics for Automatic Facial Composite Generation

The aim of this project is to experiment with different interactive nature-inspired heuristics to automatically generate a target picture through determining the vector of parameter values of an active appearance model (AAM). The AAM forms a statistical model of the human face from an input database of face images. Using this model, it is possible to generate faces which may not even exist in the original database. The natureinspired heuristics used in this study are genetic algorithms, evolutionary strategies, particle swarm optimization and differential evolution. In the interactive versions of these heuristics, users get involved in the algorithm in either the fitness evaluation or the selection stages.