基于改进YOLO v7的血细胞识别计数研究

Research on Blood Cell Recognitionand Counting Based on ImprovedYOLO v7

血细胞计数具有重要医学意义。而由于传统计数方法存在一定局限性,本文引入改进YOLO v7,对血细胞可实行快速且准确的识别计数。改进点为以下三方面:1) 更改SPPCSPC中Maxpool的尺寸。2) 将第24、37层的卷积层替换为CA注意力。3) 设计了一种上采样整体结构。改进YOLO v7在mAP方面优于原版YOLO v7,为工程应用带来一定研究意义。

Blood cell count has important medical significance. However, due to the limitations of traditional counting methods, this paper introduces the improved YOLO v7, which can quickly and accurately identify and count blood cells. The improvement points are as follows: 1) Change the size of Maxpool in SPPCSPC. 2) Replace the convolution layer of layers 24 and 37 with CA attention. 3) An upsam-pling architecture is designed. The improved YOLO v7 is better than the original YOLO v7 in terms of mAP, which brings certain research significance for engineering application.