Transfer learning based Covid-19 detection using Radiography Dataset

COVID-19 is a worldwide pandemic that affected health care and lifestyle all over the world and early discovery is critical for controlling virus spread and mortality. The principal diagnostic test is the RT -PCR test, while test results are pending, spotting probable COVID19 infections on a chest X-ray may aid in restricting high-risk individuals if the diagnosis is done early. Most medical systems have X-Ray equipment, and since most modern X-Ray systems are automated, there is no need to transfer samples. Therefore, we recommend building a Deep Convolutional Neural Network-based technique that works on Radiography images for identifying COVID-19 positive patients. Here we have applied transfer-learning over some widely used deep CNN models like NasNet, DenseNet121, VGG19, ResNet50, and Xception. We have compared the performance of each model by running them over the COVID19 Radiography Dataset. Around 40K chest X-ray photos of COVID patients were used to build training, test, and validation sets. Since earlier research, there has been significant improvement in the number of data points to better train the CNN models. This study aims to identify the best of the available solutions that can be used by medical staff to swiftly discover COVID positive persons by just using the patient's Chest X-Ray diagnosis.