Covid-19 and Pneumonia Classification using Deep learning

ABSTRACT

In order to be detected early, the COVID-19 pandemic brought on by the SARS-CoV-2 virus requires excellent diagnostic technologies. Images from chest X-rays have become a useful tool for the early detection of COVID-19. The purpose of this study is to compare how well ResNet50 and DenseNet121, two well-known deep learning models, perform when used to identify COVID-19 and pneumonia from chest X-ray pictures. Modern convolution neural network architectures ResNet50 and DenseNet121 are renowned for their outstanding performance in image identification applications. These models can successfully learn complicated characteristics from X-ray pictures since they have already been trained on the extensive ImageNet dataset. For training and testing, chest X-ray images from COVID-19, pneumonia, and normal cases were employed. To increase the models robustness and generalizability, the dataset was split into training and test sets and subjected to the proper data augmentation procedures. A classification head was added to the pre-trained base models for ResNet50 and DenseNet121 in order to improve both models. Adam optimizer and categorical cross-entropy loss function were used to train the models. Accuracy and other important criteria were used to assess the models performance. According to the results, ResNet50 and DenseNet121 were both able to identify COVID-19 and pneumonia cases from typical instances in chest X-ray pictures with promising accuracy rates. The models showed that they could pick up distinguishing traits and make precise predictions.

FULL TEXT