Emotion recognition by inclusion of age and gender parameters with a novel hierarchical approach using deep learning

In this paper, the emotion, gender and age recognition has been dealt in real time video stream. This model implements Squeezenet and mini-Xception architectures that are combined in a hierarchical order. In order to train these different networks, different types of large labelled datasets have been utilized which are available publicly through a semi-supervised pipeline to reduce the annotations efforts and time. The results show that the emotion-based system gave better performance than the previous model-based system.