Age, Gender and Ethnicity Classification from Face Images with CNN-Based Features

The problem of estimating age, gender and ethnicity from human images is encountered in many areas today, where human-computer interaction has increased considerably. These problems can be quite challenging since labels containing personal data are not found in many data sets. By using the feature maps extracted with the convolutional neural networks (CNN) model we prepared in this study, hybrid models were created with CNN, support vector machines (SVM), random forests (RF), and eXtreme gradient boosting (XGBoost) algorithms; Age, gender, and ethnicity classifications were made and the results were compared. In the experimental results, close results were obtained from the hybrid models and it was observed that they outperformed the pure CNN classifier.