Optimization on Yolov5 to Improve Accuracy for Classification of White Blood Cells

In order to solve the problem of proper classification and identification of white blood cells with an improved efficiency, this paper proposes a framework for its classification and identification based on the YOLOv5 network. Presently, determining a subtype of blood cells takes a longer time with frequent errors. The data present in the blood cell picture sets are often quite unreliable, and there are just a few publicly available dataset. The identification and analysis of WBC by a computer generally avoids human errors and cuts the time to separate the WBC to half. For the past few years, a number of in-depth research strategies for the differentiation of WBC in the image of a blood cell have been developed. The proposed research suggests that the YOLOv5 capture method can be used to find and separate white blood cells using bounding boxes. The proposed task resulted in an extraction of a WBC with a 96.3% accuracy while dividing it to 88% accuracy for a thorough diagnostic test.