LG-Yolov8: A New Method for Object Detection of Kidney Stone Images
Weilin Huang, Yitong Chen, Zhu Xiao-lin · 2024
The traditional kidney stone detection model has low accuracy and slow processing speed. In order to solve the above problems, we designed a new kidney stone detection model based on the original Yolov8, which we call LG-Yolov8 and named "Localize and Gather Yolov8". LG-Yolov8 is a new kidney stone detection model, which aims to solve the problems of low accuracy and slow processing speed of traditional models. This model is improved on the basis of yolov8. It improves the detection performance by accurately locating the kidney stone target and aggregating the information of different feature layers. In order to improve the defects of Yolov8 model in small object detection, LG-Yolov8 adds an auxiliary training head to the detection head to predict different feature layers at the same time. In addition, it uses the WloU loss function instead of the CIoU loss function to improve the overall performance. By using the public kidney stone datasets for model training, the accuracy of LG-Yolov8 is improved by 4.4%, and the mAP is improved by 2.3%