Detection of Improperly Worn Face Masks using Deep Learning – A Preventive Measure Against the Spread of COVID-19

Coronavirus disease 2019 has had a pressing impact on people all around the world. Ceasing the spread of this infectious disease is the urgent need of the hour. A vital method of protection against the virus is wearing masks in public areas. Not merely wearing masks but wearing masks properly can ensure that the respiratory droplets do not get transmitted to other people. In this paper, we have proposed a deep learning-based model, which can be used to detect people who are not wearing their face masks properly. A convolutional neural network model based on the concept of transfer learning is trained on a self-made dataset of images and implemented with light-weighted neural network called MobileNetV2 for mobile architectures. OpenCV is used with Caffe framework to detect faces in an input frame which are further forwarded to our trained convolutional neural network for classification. The method has been implemented on various input images and classification results have been obtained for the same. The experimental results show that the proposed model achieves a testing accuracy and training accuracy of 93.58% and 92.27% respectively. Optimal results with high confidence scores and correct classification have also been achieved when the proposed model was tested on individual input images. © 2022, Universidad Internacional de la Rioja. All rights reserved.

Diagnosis of Covid-19 using Deep Learning

In the global health disaster of the Coronavirus infection-2019 (Covid-19) pandemic, the health sector is avidly seeking new technologies and strategies to detect and manage the spread of the Coronavirus outbreak. Artificial Intelligence (AI) is currently one of the most essential aspects of global technology since it can track and monitor the rate at which the Coronavirus develops as well as determines the danger and severity of Coronavirus patients. In this paper, we have proposed a two-stage end-to-end Deep Learning (DL) model which can be used to predict the presence and severity of Covid-19 infection in a patient as early and accurately as possible so that the spread of this viral infection can be slowed down. Hence, based on the Computed Tomography (CT) scans or chest X-rays provided by the user as an input, the DL models are built that can forecast the presence of Covid-19 in that respective patient accurately and efficiently. In this paper, 5 DL models i.e., VGG16, InceptionV3, Xception, ResNet50, and Convolution Neural Networks (CNN) are built and their comparative analysis is carried out for the diagnosis of Covid-19. On the Google Colab GPU, the models are trained for 100 epochs on a total of 1686 images of chest X-rays and CT scans. The experimental results show that out of all these models, the model based on the Xception algorithm is the most accurate one in determining the presence of the disease and provides an accuracy of 81% and 89% on CT scans and Chest x-rays respectively. © 2022 ACM.