Facial Emotion Recognition using Deep Learning with Real-Time Convolutional Neural Networks

Abstract

Facial Expression Recognition while participating online video lessons, is a cutting-edge approach of gathering feedback from live facial expressions. It has been proposed that all educational institutions make the technology available, which would help to improve the quality of instruction by automatically monitoring the students' moods throughout the lecture and providing the lecturer with immediate feedback. This study describes a program that concurrently gauges the learner's emotional state and captures on-screen activities in real time. This work consists of two main parts: facial expression recognition is the first section of this thesis. The FER2013 data set is used to train the convolutional neural network (CNN) for this section. The system accurately categorizes seven different emotions after being fed a series of face photos from the previous stage. In second part, analyze those facial expressions of students. We created an application utilizing the open-source JavaScript face-CNN-based API's trained Facial Expression Recognition Model, and we experimentally tested it to see if it could recognize facial expressions. The aim of this study is to help alleviate disruptions to learning and instruction caused by the pandemic by presenting an intuitive way to measure concentration, understanding, and engagement expected of a productive classroom.