Research on Remote Intelligent Monitoring System for Online Examination Based on Gaze Direction Classification
Zuhui Hu, Yaguang Jing, Guoqing Wu · 2023
Aiming at the problem that online examination is difficult to monitor, a remote intelligent monitoring system for online examination based on gaze direction classification is proposed. In the experimental environment, based on the remote online examination monitoring situation, the dataset with nine types of gaze directions for multi-classification problems is established, and the image classification model is used for training. The same dataset is used for comparative experiments on the ConvNeXt network and the ResNet network, and finally the ConvNeXt network is selected as the backbone network. The ConvNeXt network model trained based on nine types of gaze directions classification dataset performs well in the weighted accuracy and weighted recall rate on the test set, with the weighted accuracy rate reaching 88.82% and the weighted recall rate reaching 86.96%. The experimental results show that the solution of a remote online examination monitoring system based on the ConvNeXt network is feasible. It can not only improve the remote online examination monitoring mode, but also have a low economic cost, which is of great significance to promote the popularization of remote online examination.