COVID-19Net: An Effective and Robust Approach for Covid-19 Detection Using Ensemble of ConvNet-24 and Customized Pre-trained Models

The Coronavirus is extremely harmful to human lungs, and hence early detection is critical to prevent the virus from spreading. However, the cumulative workflow in the routine diagnostic approach is challenging. Although chest radiographic image analysis is found to be a good alternative screening method, manual examination of abnormalities in those images requires a skilled expert. Moreover, it is a time-consuming process. The recent advancements in deep learning techniques makes them as a promising choice for the development of sophisticated applications that can meet clinical accuracy requirements. Motivated by this, a novel ConvNet-24 is proposed for efficacious classification of Covid-19 from X-ray images. Furthermore, several state-of-the-art pre-trained models (such as Alexnet, Densenet-201, Mobilenet-v2, Googlenet, Squeezenet, Inception-v3, Resnet-18, NasnetMobile, Resnet-50, Darknet-19, Resnet-101, Darknet-53, and Xception) are tailored using transfer learning technique. A novel ensemble model is proposed by investigating the models in 126 configurations, thereby improving the overall performance. Experimental findings reveal that aggregating the best models results in an overall classification accuracy of 98.5%, outperforming state-of-the-art techniques.