Depression Detection Using Convolutional Neural Networks

Abstract

Facial Expression Recognition (FER) has been an essential field in the Deep Learning area since the arrival of big data. Human psychological activity is reflected by human facial expressions that can provide significant knowledge about human nature. In that time, depression has become a severe mental illness. It is vital to detect depressed people. To detect depression, we have proposed a system based on CNN, OpenCV, Haar Cascade Classifier. Haar Cascade Classifier is a machine learning algorithm used for face detection. Convolution layers are used in the proposed combination technique. Furthermore, no datasets available that represents the natural expressions of the depressed face and can be used to detect depression. So, We have generated a Depressed and Not Depressed (DND) dataset which contains the 5000 images. Our proposed system has been evaluated using the DND dataset and achieved accuracy 81% , precision 87% , recall 88%.