Coronavirus (COVID-19) disease is spreading rapidly and is becoming increasingly common every day. This study can be helpful in the basic steps for doctors to diagnose the diseases and treat the patient accordingly. We used 1200 images of COVID-19,1300 Normal images and 1345 Viral Pneumonia images for the research work. We compared three popular deep learning methods which were CNN, VGG19, and Resnet-18 Model. We found that the Resnet-18 model with different useful parameters outperforms all the other proposed models. The accuracy we got was 97% when 50 images from each category were being tested.