ABSTRACT
The COVID-19 Omicron and Delta variants are the most prevalent sources of lethality. The computed tomography (CT) scan images may be used mostly to make an appropriate diagnosis of COVID-19. The investigation and evaluation of how well convolutional neural networks (CNNs) transfer learning methods operate in the diagnostic process was the major purpose of this research. They are capable of being trained to find anomalies in CT scan images that point to the existence of COVID-19 Omicron and Delta variants. Improved sequential CNN architecture accomplished an accuracy of 98.41%, a 0.95 F1 score, 98.76% sensitivity, and 97.75% specificity in contrast to VGG16, ResNet50, Inceptionv3, and Densenet-121.
