Extraction of facial features using genetic cellular neural networks

Abstract

This paper presents proper CNN templates to locate face region consisting of eyes,eyebrows, nose and mouth in frontal face images. Training of CNN is done by using genetic algorithms. At the begining, 72 random genes are selected. Each gene includes binary codes of the elements of the CNN templates. CNN templates are designed as symmetric to satisfy the stability. After the training process CNN templates are applied to various images.