Table 12.

Performance results comparison among the suggested model and the other previous state-of-the-art works.

Author	Architecture	Number of Images	Class	Accuracy (%)
Khan et al. [57]	CoroNet	297 COVID-19, 330 bacterial pneumonia, 310 normal, 327 viral pneumonia images.	4	89.6
			3	95
			2	99
Mahmud et al. [64]	CovXNet	305 COVID-19, 305 bacterial pneumonia, 305 normal, 305 viral pneumonia images.	4	90.3
			3	89.6
			2	94.7
Ammar et al. [52]	6 pre-trained models	150 COVID-19, 150 normal, 150 pneumonia images.	3	91.28
Mousavi Z et al. [46]	Developed LSTM network	800 COVID-19, 942 viral pneumonia, 939 healthy cases images.	3	90.0
Arsenovic et al. [60]	ResNetCOVID-19	434 COVID-19, 1100 normal, 1100 bacterial pneumonia.	3	94.1
Hemdan et al. [79]	COVIDXNet	25 COVID-19 and   25 normal images.	2	90
Sethy et al. [58]	ResNet50 plus SVM	25 COVID-19 and 25 non-COVID-19.	2	95.38
Proposed model (Dataset-1)	Tuned ResNet50V2	1143 COVID-19, 1150 viral pneumonia, 1150 bacterial pneumonia, 1150 normal images.	4	89.76
			3	97.22
			2	99.13
Proposed model (Dataset-2)	Tuned ResNet50V2	1143 COVID-19, 1150 viral pneumonia, 1150 bacterial pneumonia, 1150 normal images.	4	99.46
Proposed model (Dataset-3)	Tuned ResNet50V2	1143 COVID-19,  1150 adult pneumonia.	2	98.26