From November 26, a system that shows the congestion status of the Learning Commons on the 6th floor of the Umeda Campus is in operation, and anyone can check it from the web application.
This system was developed by third-year students of the Department of Robotics and Design Engineering during their class. The members of the group that developed the system often use vacant classrooms for self-study and lunch in their daily student life, but the classrooms were always crowded and they had a hard time finding an empty one. They thought that if they had an app that could show the crowdedness of the classrooms in real time, they could find a vacant classroom with no trouble, so they devised this system.
The system uses a high-resolution wide-angle camera to capture images of the entire Learning Commons and analyzes the captured images with YOLOv3, a machine learning model that can detect objects. The number of people detected by the analysis can be viewed on a web application created with GAS (Google Apps Script). The key point is that analyzing the camera images as they were would have placed a heavy burden on the equipment and would not have worked as expected, so we divided the images into 16 sections and analyzed each section to reduce the burden on the equipment and improve the accuracy of the analysis.
Click here for the web application (provided in Japanese)