This paper presents two systems having a novel fitness function for eyes detection using genetic algorithm (GA). The first system specifies the location of the eyes in the middle frame manually, while, the other system can specify the eyes automatically. The proposed systems are applied on two databases. The first database contains a set of images for one person. The second contains other set of images, one for each person. The second system has the ability to deal with both databases, and the other system can only deal with the first database. The proposed chromosome of candidate template on the eye consists of left top coordinates, rotate angle and size. The performance of the first system is 99.2, while, for the other system with the two databases are 97.62% and 96.82% respectively. The experiment results show that the two systems get the highest performance when the population contains 1000 chromosomes.