Monitor Pupils' Attention by Image Super-Resolution and Anomaly Detection
Valentin Rothoft, Jiaxin Si, Fei Jiang, Ruimin Shen · 2017 International Conference on Computer Systems, Electronics and Control (ICCSEC) · 2017
Academic success at young age substantially lays on in-class behavior. Looking at the right spatial element - teacher or board for instance - can be considered as the first and main step towards mental focus. In this paper, we propose a novel method for monitoring pupils' attention from classic surveillance videos, based on focus points determination and classification. This approach raises two main challenges: the low-resolution of children's faces and the categorization of the focus points itself. Indeed, the gaze point of each student is deduced from his facial landmarks, which allow us to know his head pose (by solving the PnP problem). But the resolutions of the students' faces are usually 40x40 or even smaller. Therefore, we propose to combine the super-resolution algorithm to more accurately estimate the facial landmarks of the pupils in the last few rows. To achieve the second hard task - classification - and detect the children who do not pay attention to the proper spatial area, we consider the distribution of the focus points in two dimensions. After what we discriminate the anomalous points on a density criterion. Experiments on several real videos demonstrate the effectiveness and efficiency of this process.