Real Time Emotion Detection of Humans Using Mini-Xception Algorithm

Abstract

In the past few decades there has been operating analysis on emotion expression recognition due to the great intra-class deviation also it is still challenging. Maximum number of research work performs the best on controlled datasets (i.e., small datasets with less features), whereas it fails to operate well and it’s still challenging on datasets varies variations in images and even in partial faces. In modern years, many works have introduced an end-to-end plan for emotion expression recognition, utilizing deep learning models. Although emotion recognition is a great task, there still seems to be a huge area for development. In this paper, we developed a mini-Xception based on Xception and Convolution Neural Network (CNN), which is easy to concentrate on great parts like the face, and conclude important improvements to earlier works. We validated our model by creating a real-time vision system which accomplishes the tasks of face detection, and emotion classification simultaneously in one blended step using our proposed mini-Xception architecture. We still utilize a visualization technique that is ready to detect important face sectors because recognizing various emotions, based on the classifier’s output. For experimental analysis we had used FER-2013 dataset and results manifest that the proposed method can efficiently perform all the tasks like detection and classification with seven different emotions using with Mini-Xception algorithm and achieved accuracy around 95.60%.