Research on Malaria Parasite Detection in Thick Blood Smears Based on YOLO-VF
Xiujie Zhang, Baiyi Chen · 2023
Aiming at the low detection efficiency of Plasmodium vivax in microscopic thick blood smears of malaria patients, which is prone to omission and false detection, a lightweight deep learning algorithm YOLO-VF is proposed, which is based on the improvement of YOLOv7-Tiny, which improves the coordinate regression loss function and up-sampling operator, and introduces a dynamic convolutional and dynamic detection head. It is experimentally demonstrated that the improved scheme provides more accurate detection of microscopic pathogens in the images and achieves superiority in detection speed. Tested on images containing three datasets of Plasmodium vivax, Plasmodium falciparum, and uninfected people, the average detection accuracy is improved by 1.9% while the model inference speed reaches 29.772ms and the model parameters are reduced by 50.76%.