COVID-19 SEVERITY PREDICTION USING DEEP LEARNING MODELS – COMPARATIVE STUDY

Abstract

In response to the exponential rise in COVID19 instances, healthcare professionals have been seeking for strict automated detection measures to stop COVID from spreading while simultaneously attempting to restrict device processing needs. Additionally, we must comprehend the severity of the COVID infection based on the harm the infection has done to the lungs to properly treat impacted individuals. We choose a suitable CNN model as a result after performing an initial comparison analysis of many well-known CNN models. Using a dataset generated from sources like prior papers and other internet resources, this study provides an ensemble model that is exact and effective and is built on deep learning Convolutional Neural Networks (CNN). The dataset is altered in several ways to boost its accuracy. To choose the optimal deep-learning model for our application, we are performing a comparative study. We assessed the benefits and drawbacks of many commercial CNN models, including VGG16, VGG19, and Densenet 121. To select the optimal model for provided dataset, the accuracy comparative study is constructed. VGG16 model had highest accuracy, coming in at 82.46 %. The study was conducted to determine the optimum way for our multi-modal picture classification, which, with a little modification, functions well, rather than to accurately assess how well each method did.