Implementation of YOLOv3 and YOLOv4 CNN Algorithm in Detecting Struvite Crystals in Canine Urine Sediments

A microscopic examination is performed by a veterinary lab personnel to detect entities present in urine sediments. Some of these entities are casts, white blood cells, red blood cells, and crystals. You Only Look Once (YOLO) is a multi-scale algorithm suitable for detecting small objects – like struvite crystals. The major objective for this study is to make a performance comparison between YOLOv3 and YOLOv4 in detecting struvite crystals in canine urine sediments. The testing had a total of 20 canine urine sediment samples. The system’s accuracy has been determined using confusion matrix. YOLOv3 had an accuracy of 77.05%, while YOLOv4 had 73.77%. A ttest was used to compare the struvite crystals detected by the two versions, and the null hypothesis was not rejected. Thus, there is no significant difference between the two versions. Ttest was also used for the processing time of the two versions, which resulted in the rejection of the null hypothesis, revealing a significant difference in the processing time, with YOLOv3 having the faster average processing time of 45.24 seconds compared to the YOLOv4’s 50.37 seconds.