Class size and student-to-teacher ratio are the two most important factors determining the quality of teaching and learning in a classroom setting. In southern Asian countries, especially India, the class batch sizes are substantially large, leading to a very high student-to-teacher ratio of approximately 60:1. While the government plans to improve the availability of teachers via various policy measures, technological support to the existing teaching community is much needed in helping them raise the standards of education in India. This project proposes an emotion detection algorithm, which can be employed in a typical Indian classroom with a high student-to-teacher ratio. Currently, the Convolutional Neural Network designed as a part of the algorithm stands at 86% accuracy. The model successfully detects 7 primary emotions—happiness, sadness, disgust, surprise, anger, fear, and neutral. These are mapped to high, medium, and low engagement levels. The algorithm processes the real-time images of the students in a classroom using Facial Emotion Recognition (FER). It determines the emotions and then maps them to the appropriate engagement levels. The project has valuable implications for the teaching community. The teachers will be able to see students’ engagement reports class wise, or weekly/monthly, helping them identify student engagement trends, and employ appropriate interventions to improve student engagement and learning outcomes.