COVID-19 Diagnosis Using Chest X-ray Images via Classification and Object Detection

Abstract

We diagnose the symptoms and the localization of the affected area in COVID-19 cases/patients using chest x-ray images provided by the Kaggle competition. By training and predicting symptoms and the localization of the affected area using the YOLOv5 object detection algorithm, we obtained a low accuracy of approximately 20%. However, we improved the accuracy to approximately 80% by using the image classification model Keras / EfficientNetB7, in addition to YOLOv5. Although it is difficult to detect visually ambiguous objects such as pneumonia, we believe that we can improve the accuracy by training/predicting symptoms using the image classification model and the localization of the affected area using the object detection algorithm.