Blood Cell Detection Method Based on Improved YOLOv5
Yecai Guo, Mengyao Zhang
Abstract
In order to solve the problems of low accuracy and missed detection in traditional blood cell data detection tasks. This paper proposes and implements the blood cell detection method based on the YOLOv5 (YOLOv5-ALT). The goal of this research is to enhance the accuracy of the detection with the YOLO techniques. This work presents the method overcomes the shortcomings of the existing method by introducing the attention mechanism in the feature channel, modifying SPP module in YOLOv5 backbone feature extraction network, and changing the bounding box regression loss function. Based on the deep learning object detection algorithm, each evaluation index is compared to evaluate the effectiveness of the model. Experimental results show that the [email protected], Precision and Recall of the YOLOv5-ALT reaches 97.4%, 97.9% and 93.5%. This method is more in line with the effectiveness of the blood cell detection task.