A CNN-Based Approach for Facial Emotion Detection

One of the most versatile ways in which individuals express their state of mind is through facial expressions. The advancement of deep learning-based technologies helped us to detect human emotion from images that can be used for understanding human feelings as well. The image can be static or can be captured through a web camera in real time. The precise analysis of human facial expressions is necessary for a better understanding of human behaviour. With the recent progress in deep learning, Convolution Neural Network (CNN) with its enhanced complex architecture is capable of emotion detection in a much better and more efficient way. In this paper, we experiment and demonstrate how to build a CNN predictor model using TensorFlow that can predict the emotion from images of human facial expressions with satisfactory accuracy. Additionally, we also develop an application that asks for image input from the user and predicts the emotion from the given input image. Through this experiment, we are successful in demonstrating how CNN is an appropriate model for this task. Our work is beneficial in many applications such as lie detectors and student assessments to detect facial expressions very accurately.