Forecasting COVID-19 via Registration Slips of Patients using ResNet-101 and Performance Analysis and Comparison of Prediction for COVID-19 using Faster R-CNN, Mask R-CNN, and ResNet-50

Abstract

Covid-19 was an outbreak of unfamiliar diseases affecting the majorly respiratory system. The disease gradually exacerbates and inscribed the whole world. Majorly deteriorating population. The well versed and optimized methodology to predict Covid-19 is still questionable. However, state of the art techniques of Deep Learning clearly created a new path of superior prediction and forecasting. This study consists of two dimensions. First Dimension is that a new method of prediction has been established via COVID-19 patients registration slips. ResNet-101 has been applied to indigenous data set of COVID-19 patients registration slips. The dataset includes 5003 E-registration slips with exact timings. The accuracy of forecasting with respect to time was 82%. Forecasting for COVID-19 positive cases of the following day was determined. Error framework was also established inscribing of MOE, MAE juxtaposition. The second Dimension includes the prediction of COVID-19 via Chest X-Ray. Indigenous dataset of 8009 chest X-Ray was collected. Three Neural Networks were implied incriminating Faster R-CNN, Mask-CNN, and ResNet-50. Faster R-CNN shows the best accuracy of 87%. Mask R-CNN accuracy was 83% and resNet-50 ended up at 72%. Performance parameters were in terms of ACC, PRC, and RCL. The batch normalization technique was added to improve SVM performance..

Ähnliche Arbeiten