Application of Pruning Yolo-V4 with Center Loss in Mask Wearing Recognition for Gymnasiums and Sports Grounds of Colleges and Universities
Xiaoyu Wang, Zhiyong Chen, Bohan Wei, Maotian Ling · 2020
Since the first case of COVID-19 was founded, the international organization began to fight against the epidemic. In addition to various research and findings from experts in medical and health care, the daily behavior of citizens has became the key of fighting against the epidemic. In China, thanks to the government's active measures and citizens' spontaneous compliance with policies such as quarantine and wearing masks when going out. As the first country be attacked, China becomes the country with the best epidemic-control in the world. However, it is not enough for the masses to wear masks consciously by citizens, and it is still necessary to supervise the wearing of masks in various public places by governments. In this process, this paper proposes to replace manual inspection with a deep learning method, using the most powerful object detection algorithm YoloV4, supplemented by CenterLoss and pruning algorithm, so that it can be better used in the actual environment, especially regarding the supervision of wearing masks in indoor stadiums and college physical education classes. The experimental results prove that the algorithm in this paper can effectively deal with multiple targets, occlusions, different sizes of people, and achieve effective supervision of personnel.