An Upscalable Distributed Deep Learning based System for Multi-class Xray Covid19 Classification

Abstract

In this paper, we perform the problem of X-ray image classification between Covid infected lungs and normal lungs by using Apache Spark. Our system deploys deep learning models on Spark framework by using BigDL library in order to handle a big data of chest X-ray images with high performance. The deep learning models used are CNN, Resnet50, VGG90, InceptionV3 trained on a big data of more than 13.000 chest X-ray images for COVID-19 positive cases along with Normal cases. We achieved pretty good results with all deep learning models, in which the InceptionV3 model gave the highest performance of 93.75% with accuracy measurement.