Transfer Learning from Pneumonia to COVID-19

摘要

Developing an intelligent application to assist the detection and study of the COVID-19 infection is crucial and urgent during this pandemic, given the scarcity of available data and the rapidly changing virus. This paper presents a study of transfer learning in image classification to efficiently develop deep learning models following a three-stage procedure to take advantage of pre-trained models from one area and effectively modify the model for application in a relatively new area. The case study in this work is the classification of COVID-19 X-ray images. The experiment evaluations show that the proposed method and developed models achieve satisfactory results in a timely manner.