TABLE II. Comparison of Various Lightweight Deep Learning Models Utilized in This Work in Terms of Parameters, Corresponding Required Memory (in MB), as Well as the Number of FLOPs.

Method	Total no. of Parameters	Trainable	Non-Trainable	Memory (in MB)	Giga Flops
COVID-CAPS	295,616	295,488	128	4.51	8.8
COVID-CAPS(Focal)	295,616	295,488	128	4.51	8.8
COVID-CAPS Scaled	3,094,432	3,093,920	512	37.82	55.4
COVID-CAPS Scaled (Focal)	3,094,432	3,093,920	512	37.82	55.4
POCOVID-Net	14,747,971	2,392,963	12,355,008	225.04	30.7
POCOVID-Net(Focal)	14,747,971	2,392,963	12,355,008	225.04	30.7
Mini-COVIDNet	3,361,091	3,338,947	22,144	51.29	1.15
Mini-COVIDNet(Focal)	3,361,091	3,338,947	22,144	51.29	1.15
MOBILE-Net-V2	2,422,979	2,388,611	34,368	36.97	0.613
MOBILE-Net-V2(Focal)	2,422,979	2,388,611	34,368	36.97	0.613
NASNetMOBILE	4,406,039	4,369,045	36,994	67.23	1.15
NASNetMOBILE(Focal)	4,406,039	4,369,045	36,994	67.23	1.15
ResNet50	23,851,011	23,797,635	53,376	291.30	7.75
ResNet50(Focal)	23,851,011	23,797,635	53,376	291.30	7.75