• This program calculates the attention span of students in online classrooms based on gaze tracking. • Input data from webcam - which after processing into text data using OpenCV & Pupil Detection - is fed into a simple neural network. • Pupil tracking was done using the Tensorflow Object Detection API. • 10 other features were also extracted with the help of simple computer vision techniques by using OpenCV & Dlib library. • Gaze tracking is successful even without requiring to initialize for the first time. • Makes 6 prediction per second. It is faster than many current models since it uses a text based simple neural network rather than a full blown CNN.
• This program calculates the attention span of students in online classrooms based on gaze tracking. • Input data from webcam - which after processing into text data using OpenCV & Pupil Detection - is fed into a simple neural network. • Pupil tracking was done using the Tensorflow Object Detection API. • 10 other features were also extracted with the help of simple computer vision techniques by using OpenCV & Dlib library. • Gaze tracking is successful even without requiring to initialize for the first time. • Makes 6 prediction per second. It is faster than many current models since it uses a text based simple neural network rather than a full blown CNN.