This project focuses on analyzing urine cells in the medical imaging domain. Utilizing a dataset obtained from microscopic images, it aims to provide real-time detection, classification, and health status information regarding urine cells using the YOLO algorithm. The YOLO algorithm stands out as a fast and accurate method for object detection, being the preferred method in this project.
The detected cells include Bacteria, Other, Epithelial, Erythrocyte, Crystal, Leukocyte, Yeast, and Cylinder. Evaluations were conducted on YOLO algorithm versions yolov6, yolov7, and yolov8, with the best performance achieved by the yolov8 large version, reaching an 85% success rate.
Numerical distribution of microscopic images of urine cells in the dataset
Confusion Matrix
F1 Curve
PR Curve
