EAGCNN-CNN based Combined Model for Human Emotion, Age and Gender Estimation

Abstract

Demand of computer vision-based applications in daily life scenario has grown drastically such as object detection, recognition and classification. Currently, emotion recognition is considered as a challenging and important task in various applications. Several techniques have been introduced but due to poor lighting conditions, illumination, occlusion and aligned faces, the existing approaches fail to achieve the desired results. Recently, CNN (Convolutional Neural Network) has gained attraction in research field. In this work, we use CNN model and introduce a combined model for emotion classification, age and gender estimation with the help of CNN and Caffe models. The proposed CNN architecture shows better performance for offline dataset and for real-time experiment.