Interval Valued Data Representation for Gender Classification of Celebrity Cartoon Faces

In this paper, we address the problem of recognizing the gender of celebrities in cartoons images. We propose a method of representing celebrity cartoon faces by the use of interval-valued features and suitable symbolic classifier for gender classification. The FaceNet architecture which is recommended for Face recognition is adopted to extract features for cartoon faces. In each class, the similar looking celebrity cartoon images are clustered using K-means clustering algorithm. The approach is carried forward to preserve the intra-class variation present in each cluster corresponding to each class using a vector of interval-valued data is empirically evident. Thus, a cluster of gender cartoon faces is represented in the form of a vector of intervals. Further, a method of reducing the dimensionality of interval-valued features is employed to reduce considerably the number of interval-valued features for compact representation to have efficient yet effective classification with a minimal number of feature set by reducing computational burden is add on in our proposed methodology than the contemporary models. However, based on gender celebrity cartoons identification and facial expression prediction is kept beyond the scope of this paper as it is our future target. For experimentation purpose, we have used IIIT-CFW Celebrity Cartoon Face database. The results show that the proposed model outperforms the other existing models with respect to F-measure.