Detection and classification of white blood cells with an improved deep learning-based approach

The analysis of white blood cells, which defend the body against deadly infections and disease-causing substances, is an important issue in medical world. concentrations these cells blood, examined 5 classes, i.e. monocytes, eosinophils, basophils, lymphocytes, neutrophils, vary according to types diseases body. peripheral smear widely used analyze cells. Manual evaluation this method laborious time-consuming. At same time, many environmental humanistic parameters affect method's performance. Therefore, presented study, a real-time detection process realized. Firstly, YOLOv5s, YOLOv5x, Detectron 2 R50-FPN pretrained models object recognition framework are used. Next, two original contributions made study improve model's first contribution includes optimizing activation function, essential criterion training model, arrangement provided architecture. With proposed approach, improvement 0.006 achieved rates all classes. second combined use YOLO Detectron2 frameworks, have different processes. success rate with hybrid structure between 3.44% 14.7% compared outputs obtained from models. In addition, maximum accuracy on test dataset for classification as 98%.