Feasibility Investigation for Estimation of Audiences' Attention with Semi-supervised Learning

Abstract

An automated estimation of a degree of audience's attention might be helpful not only for a specific feedback for improving education methods but also one of indicators of the lecturing skill.The estimation method of audience's attention with Convolutional Neural Networks (CNN) has achieved the state-of-the-art result under actual environment.It demands, however, a huge task in assigning the training labels beforehand in exchange for its accuracy.This is one of the main difficulties for introducing this methodology into the practical use. In this paper, we introduced an active learning and semi-supervised learning to reduce the troublesome labelling tasks for the abovementioned purpose and reported the latest results.