Table 5. Prediction performance of proposed ECOVNet without using ensemble.

Method	Pre-trained Weight	Precision(%)	Recall(%)	F1-score(%)	Accuracy(%)(95% CI)
ECOVNet (Without Augmentation)	EfficientNet-B0	93.27	93.29	93.27	93.29± 1.23
	EfficientNet-B1	94.28	94.30	94.26	94.30± 1.14
	EfficientNet-B2	93.24	93.03	93.08	93.03± 1.26
	EfficientNet-B3	95.56	95.57	95.56	95.57± 1.01
	EfficientNet-B4	95.52	95.50	95.50	95.50± 1.02
	EfficientNet-B5	96.28	96.26	96.26	96.26 ± 0.94
ECOVNet (With Augmentation)	EfficientNet-B0	91.71	74.10	79.72	74.10± 2.16
	EfficientNet-B1	91.02	86.19	87.67	86.19± 1.70
	EfficientNet-B2	93.60	93.10	93.24	93.10± 1.25
	EfficientNet-B3	92.60	90.25	90.92	90.25± 1.46
	EfficientNet-B4	94.32	93.73	93.89	93.73± 1.20
	EfficientNet-B5	94.79	94.68	94.70	94.68 ± 1.11

Note:

Bold indicates that the method has statistically better performance than other methods.