Optimizing Deep Convolutional Neural Network With Fine-Tuning and Data Augmentation For Covid-19 Prediction

This is our paper which was accepted and published at IEEE International Conference on Internet of Things and Intelligence Systems (IoTaIS), 2021, pp. 169-175, DOI: 10.1109/IoTaIS53735.2021.9628799.

Abstract — Since Corona virus disease 2019 (Covid-19) has been infecting people worldwide, it is important to detect Covid19 at an earlier phase to fight against the pandemic. Pathogenic and laboratory testing are needed to determine whether someone is infected or not by Covid-19. However, this laboratory test is relatively time consuming and could produce significant false
negative rates. This paper presents a study on Covid-19 detection by using deep learning algorithms aiming to predict and detect Covid-19. A set of chest X-ray images are used as the input datasets to prepare and to train the proposed model. In this study, a deep learning architecture (DLA) and optimisation strategies have been proposed and investigated to maintain the automated Covid-19 detection. A platform and a model model based on convolutional neural network (CNN) is introduced to extract the feature of X-ray images for feature learning phase in order to make the model suitable for the problem. Two strategies
are applied to improve the performance of proposed model, i.e. Data augmentation and fine-tuning with deep-feature-based. A classifier are employed in order to enhance the performance of model. The experimental investigation was performed between the proposed work with the pre-trained DLAs, such as VGG16 and ResNet50. The results of this study affirm that the proposed model and VGG16 obtain better classification accuracy of 98%
and 95% of sensitivity respectively.


Index Terms—Deep Learning, Convolutional Neural Network, Fine-Tuning, Data Augmentation, Covid-19, Detection, Prediction

You can download the paper here.

Leave a Reply

Your email address will not be published. Required fields are marked *