Table 8.

Comprehensive performance comparison of all evaluated models

Model Category	Model	Accuracy	Macro Precision	Macro Recall	Macro F1-Score
Custom Architecture	CNN	0.743 ± 0.000	0.743 ± 0.000	0.703 ± 0.000	0.724 ± 0.000
Hybrid Models	CNN + DT	0.812 ± 0.029	0.802 ± 0.023	0.795 ± 0.025	0.798 ± 0.026
	CNN + RF	0.854 ± 0.023	0.847 ± 0.024	0.839 ± 0.030	0.843 ± 0.028
	CNN + SVM	0.786 ± 0.031	0.763 ± 0.028	0.762 ± 0.033	0.763 ± 0.024
Pre-trained Models	VGG16	0.823 ± 0.020	0.816 ± 0.027	0.818 ± 0.030	0.817 ± 0.027
	Xception	0.809 ± 0.023	0.800 ± 0.027	0.801 ± 0.028	0.800 ± 0.027
	ResNet50	0.795 ± 0.027	0.783 ± 0.023	0.789 ± 0.028	0.786 ± 0.026