NOVEL APPROACH FOR GENDER CLASSIFICATION IN ANIMATED MOVIES

Abstract

Gender of fictional characters in a computer graphic animated film was identified focusing on face using LBP histogram as well as Gabor filtering. Fundamental gender classification steps of image acquisition including normalization and feature extraction along with two classification tools (ANN and SVM) were used separately. The potential of proposed feature extraction method was proved through 180 still images of fictional characters for training and four animated movies utilized for Testing of more than one hour duration each. The accuracy was measured 87.5 to 100 percent under applied setup. Results revealed high degree of agreement with theoretical predictions and significant improvement over previous efforts on still face photos and conventional camera movies.