Table 2.

Comparative table of object-detection CNN models’ performance.

CNN model	Validation dataset				Test dataset
	Precision	Recall	F-score	mAP 0.5	Precision	Recall	F-score	mAP0.5	p-value
YOLOv5x	0.8975	0.9197	0.9085	0.9490	0.9210	0.9350	0.9279	0.9440	0.157
Faster R-CNN	0.8753	0.9331	0.9033	0.9194	0.8913	0.9638	0.9261	0.9412	0.144
SSD	0.9501	0.4789	0.6368	0.8491	0.9562	0.5599	0.7063	0.9133	0.354
RetinaNet	0.9369	0.8155	0.8720	0.9180	0.9407	0.8719	0.9050	0.9489	0.187

Descriptive parameter values of Precision, Recall, F-score, and Mean Average Precision (mAP0.5) are represented for Validation and Test datasets. YOLOv5x: You Only Look Once version 5 model x, Faster R-CNN: Faster R-Convolutional Neural Network, SSD: Single Shot Detector. Statistical analysis (t-test) to compare the performance of CNN models with validation and test data subsets (p < 0.05). Bold values represent the higher values of each parameter, in validation and test datasets.