Abstract
The COVID-19 has become a huge factor affecting human health and has caused huge losses to the world. How to accurately and effectively distinguish the infected has become an important measure to prevent the spread of the epidemic. However, the currently commonly used detection technology is Polymerase chain reaction (PCR), which is very labor-intensive, dangerous and inaccurate by manual detection and differentiation by humans. Therefore, this paper uses the MobilenetV3 network model to conduct a large number of experiments in 8282 lung X-ray images, and finally achieves fast and accurate classification of four types of lung X-ray images, and the classification accuracy rate reaches 97.2%. The COVID classification precision, recall and specificity reached 100%, 98.5% and 100%, respectively.