"The system is intended to 'visualize' what children are doing, not to monitor them, and help teachers improve their teaching skills," said Kazuki Niwayama, an associate professor at the university. By comparing children's behavior with data on how they are being taught, the system can, for example, encourage teachers to compliment students more frequently, provide more simple explanations, and devise other ways to deepen understanding, the project team said. The team, led by Fumio Nakaya, a professor at the university's Center for Education Innovation Design, compiled a multitude of images of people sitting, standing, dozing off and engaging in other activities. When the team fed the images in the AI system for use in a classroom, its analysis matched the actual behaviors exhibited by students with an accuracy of more than 90 percent, the team said. It is impossible to identify individuals through thermal images, which are different from regular camera footage, and leaks of classroom thermal images to the outside should cause no problems in terms of privacy, the team said.
