Blood Cell Target Detection Based on Improved YOLOv5 Algorithm
Xuan Song, Hongyan Tang · Preprints.org · 2024
In the field of medicine, blood analysis is one of the most important means of assessing human health. The type and number of blood cells is an important basis for doctors to diagnose and treat diseases. To solve the problems of difficult classification and low efficiency in blood cell detection. In this paper, an improved YOLOv5-BS blood cell target detection algorithm is proposed. The purpose of the improvement is to improve the real-time and accuracy of blood cell type identification. The algorithm is based on the YOLOv5s network and adds the advantages of CNN and Transformer. First, combine the BotNet backbone. Then replace the YOLOv5 Head architecture SPP-YOLO with Decoupled Head architecture. Finally, a new loss function SIoU is used to improve the accuracy and efficiency of the model. In order to test the feasibility of the algorithm, a comparative experiment is done. Experiments show that the average accuracy of the improved algorithm on the test set reaches 83.3%, and the recall rate is 99%. Compared with YOLOv8, the increase was 3.9% and 3% respectively. The algorithm effectively improves the efficiency and accuracy of blood cell detection, and effectively improves the problem of blood cell detection.