Detection of white blood cells using YOLOV3 network
Gan Fang, Yang suhua, Jiang shaofeng · 2019
White blood cells are a very important part of medical research diagnosis. With the rise of artificial intelligence, it is also widely used in the medical field. At present, deep learning is the main research method. Most neural networks do not perform well on white blood cells. In this paper, we use YOLOv3 to improve the accuracy rate. In yolov3, Darknet extracts features and k-mean produces boxes. Yolov3 is Multi-scale prediction, so YOLOv3 is better for small target detection. At the same time, in this paper, we also preprocessed the data set by portioning them into blocks according to a certain ratio to improve the accuracy rate. As the results shown, the processed data set is better performance than the original data set in training.