A SVM based Gender Classification from Children Facial Images using Local Binary and Non-Binary Descriptors

Gender being an important component within the interaction between individuals. Generally in face acknowledgment systems, local descriptors are broadly used as descriptors of features to attain more appealing results in dynamic conditions of the image such as lights & shades, pose etc. The complex characteristics of human confront are sex, facial expressions, age, ethnic origin etc. In this paper, we are taking a challenge to predict the gender of kids between 6 to 8 years by analyzing there features with various methods like Local Binary Pattern (LBP), Histogram Oriented Gradients (HOG) Local Directional Pattern (LDP), and the SVM-Support Vector Machine is utilized in defining the gender, this includes three kernel functions. Exponential results indicate that combination of HOG & SVM linear kernel get better performed on the CFID which we have created of about 60 images with an accuracy of 96.66 percentage.