The Entering and Exiting Management System by Person Specification Using Deep-CNN

We developed a system to manage time when students stayed at a laboratory. In order to eliminate the cumbersome authentication operation at the time of entering and exiting a laboratory, the duration that students stayed is collected by facial detection, facial recognition using the Deep-CNN, and entering and exiting detection. We narrowed the range of the facial image, increased the speed of facial detection processing, and transformed geometrically to improve facial recognition accuracy. As a result, the success rate of the facial recognition of five target people was about 97%. Further, the entering and exiting detection rate was about 84% by insertion of an entering and exiting end and by scoring using the recognition result of images. Thus, we could confirm that our system is practical.