Research Article
Corona Virus Detection Using Transfer Learning Technique
@INPROCEEDINGS{10.4108/eai.7-12-2021.2314728, author={Sangeetha Balachandran and Senthil Prabha R and Ravitha Rajalakshmi N and Srilam K}, title={Corona Virus Detection Using Transfer Learning Technique}, proceedings={Proceedings of the First International Conference on Combinatorial and Optimization, ICCAP 2021, December 7-8 2021, Chennai, India}, publisher={EAI}, proceedings_a={ICCAP}, year={2021}, month={12}, keywords={corona virus detection transfer learning techniques mobilenet inceptionresnetv2 covid 19 detection in chest x-rays vgg19 densenet cnn}, doi={10.4108/eai.7-12-2021.2314728} }- Sangeetha Balachandran
Senthil Prabha R
Ravitha Rajalakshmi N
Srilam K
Year: 2021
Corona Virus Detection Using Transfer Learning Technique
ICCAP
EAI
DOI: 10.4108/eai.7-12-2021.2314728
Abstract
The COVID-19 pandemic has caused unprecedented upshot reporting around 234 million people affected worldwide. This alarming effect has led to the shortage of testing kits in many countries consequently having an impact on mortality. To combat this issue, other means for detecting the infected patients have to be put in place rather than relying on polymerase chain reaction (PCR). Radiographs are the easiest way to diagnose a patient with COVID-19 or not. This research work aims in developing a Computer-aided diagnostic system to assist radiologists in the detection of the COVID-19 virus. The proposed work builds a deep learning model with Convolutional Neural Network as its baseline architecture. Due to limitations in the number of the dataset, this work uses transfer learning techniques to exploit the benefit of pre-trained models and generalize it for COVID-19 detection. Experimentation is done using 4 different models like VGG 19, MobileNet, DenseNet and InceptionResNetV2. The performance of these pre-trained models is evaluated by changing the optimizers and initializers. The covid19-radiography-database dataset used for experimentation consists of 3568, 1345 and 1164 COVID-19, pneumonia and standard chest X-ray images respectively. It is observed from the experiments that the MobileNet architecture reports the highest accuracy of 99.67% using RandomUniform initializer and Adam optimizer.
