Evaluation method of children's learning concentration based on deep learning

Abstract

In view of the current young parents busy with work, it is difficult to accompany their children to study, and the corresponding supervision methods are relatively scarce. This paper proposes a method for evaluating children's concentration based on deep learning, using improved CenterNet and MobileNet-V2 network cascading to realize children's face region localization and 68 feature points extraction, while maintaining the detection speed, but also taking into account the detection accuracy. According to the extraction of the corresponding feature points, the head posture and fatigue are estimated, and then the AHP analytic method is used to calculate the corresponding weights, and finally the quantified evaluation of children's learning concentration is realized to help children's concentration training. Experiments show that the real-time and accuracy of this method can be used for special evaluation of children's learning.