A Students’ Action Recognition Database In Smart Classroom

With the development of human action recognition, it is possible to automatically recognize students' actions in classroom, providing a new direction for classroom observation in teaching research. Training effective students' action recognition algorithms depends significantly on the quality of the action database. However, only a few existing action databases focus on learning environment. In this paper, we contribute to this topic from two aspects. First, a novel students' action recognition database is introduced. The spontaneous action database consists 15 action categories, 817 video clips of 73 students, which are collected in real smart classroom environment. Second, a benchmark experiment was conducted on the database using two kinds of recognition algorithms. The best result is achieved by Inception V3 with 0.9310 accuracy. Such a spontaneous database will help in the development and validation of algorithms for action recognition in learning environment.