ABSTRACT
SARS COV-2 or Novel Coronavirus COVID19 is now the world's most difficult problem;it has turned pandemic, and a large number of people have died during this pandemic era. The mortality rate is much higher than that of any other illness in the past century. A vaccination against this virus has not yet been developed. The virus is detected via RTPCR testing, which is not 100 percent reliable, is expensive, and there is a scarcity of test kits. Thus, the objective of this approach is to identify COVID-19 utilizing both a conventional system and a deep learning method with increased accuracy and availability. This approach proposes a Convolutional Neural Networking technique with 19 consecutive layers and a dataset consisting of two classes of data: COVID and Normal x-ray. The collection contains 1621 pictures, 280 of which are of COVID patients and 1341 of which are of normal patients. While training using the suggested Convolutional Neural Network architecture, the K-fold cross validation is utilized, and the folding is performed five times. Prior to incorporating pictures, some preprocessing was performed, such as via the use of an anisotropic diffusion filter. The suggested technique is 99.5 percent accurate on average. © 2021 IEEE.