Classification of Student Attentiveness on Video Data Using Facial Expressions Extracted from Minimal Image Sequences
Term of Award
Master of Science, Computer Science (M.S.C.S.)
Document Type and Release Option
Thesis (restricted to Georgia Southern)
Copyright Statement / License for Reuse
Digital Commons@Georgia Southern License
Department of Computer Science
Committee Member 1
Gursimran Singh Walia
Committee Member 2
Attentiveness is an important indication for student success as it can demonstrate comprehension and the effectiveness of the teaching technique. To facilitate an effective learning environment, tracking the overall attentiveness of each student is very important. However, it becomes challenging for an instructor when many students are physically present in the classroom, when attendance is via video conference, or a combination of both. One way that overall attentiveness can be expressed is as a function of a temporal sequence of facial expressions. This thesis investigates deep learning models' performance on the classification of attentiveness using extracted sequential facial expressions from recorded video data and compares the results. Models are based on long short-term memory networks(LSTM) and convolutional neural networks(CNN). CNN layers are involved in feature extraction, and LSTM layers are involved in sequence prediction. We took 30 minimal image sequences from each 10-second video in the publicly available DAiSEE engagement dataset. We extracted facial emotions from images using an emotion detection algorithm. Our hybrid model(CNN-LSTM) outperformed the LSTMs only model by achieving 89.6% accuracy.
Jarman, Angur M., "Classification of Student Attentiveness on Video Data Using Facial Expressions Extracted from Minimal Image Sequences" (2021). Electronic Theses and Dissertations. 2254.
Research Data and Supplementary Material