Research on the Use of YOLOv5 Object Detection Algorithm in Mask Wearing Recognition

Yifan Liu ； BingHang Lu ； Jingyu Peng ； Zihao Zhang

《World Scientific Research Journal》 6卷11期 (2020/11) Pp. 276-284

https://doi.org/10.6911/WSRJ.202011_6(11).0038

摘要

Masks can help people to reduce inhalation of droplets and the risk of infection. Because of the COVID-19, many governments required people to wear marks to prevent virus spread. In some public places, there are tons of people going back and forth everyday so it's impossible to settle a human monitor to identify whether everyone wears a mask. This work uses a different training version from YOLOv5 to train the dataset of mask wearing, and we use K-means to find the most appropriate anchors for datasets. Finally, by using data augmentation we get a more accurate model. Compared to human work, this model can be faster and more accurate to find a target and it can save countless money and time.

關鍵字

Masks ； YOLOv5 ； Computer vision ； Machine Learning

參考文獻

R. Girshick, “Fast R-CNN,” in IEEE International Conference on Computer Vision (ICCV), 2015.

Joseph Redmon, Santosh Divvala, Ross Girshick, Ali Farhadi: “You Only Look Once: Unified, Real-Time Object Detection”, 2015; arXiv:1506.02640.

Glenn Jocher, yolov5, (2020), GitHub repository, https://github.com/ultralytics/yolov5

CSDN, Wanzhuandeepleaening, 2020, Theory and Evolution of the YOLOv4 Model, https://blog.csdn.net/shajiayu1/article/details/105755280

Joseph Redmon, Ali Farhadi: “YOLOv3: An Incremental Improvement”, 2018; arXiv:1804.02767.