Emotion Recognition From Online Classroom Videos Using Meta Learning

Emotion recognition is one of the most important application of computer vision and artificial intelligence. Academic and online teaching institutes must be able to recognize emotion of students from classroom video. This helps to determine the attitude of the students and also devise techniques to engage students that makes learning an interesting activity. This paper presents work on emotion recognition from online classroom videos using layer based Convolutional Neural Networks (CNN) and Siamese Neural Network. The proposed method for emotion recognition is named as SNSER (Siamese Network for Student Emotion Recognition Model). For training the model CAFE dataset is used and an accuracy of 80% is obtained. Neutral, Anger, Happy, Surprise, Sad, Fear, and Disgust are the emotions considered for training the model. In addition to these 7 basic emotions used during training, boring and confused are also included for testing.