Table 6.

Accuracy summary of different ResNet50 TL models during 50 epochs of training where Acmx $=$ Maximum Accuracy, Acmn $=$ Minimum Accuracy, Avacy $=$ Average Accuracy, E $=$ epoch, INCXR14 $=ImageNet\text{\_}ChestX-ray14$, INCxP $=ImageNet\text{\_}ChexPert$, iNSup $=iNat2021\text{\_}Supervised$, iNSupFS $=iNat2021\text{\_}Supervised\text{\_}From\text{\_}Scratch$, iNMSwAV $=iNat2021\text{\_}Mini\text{\_}SwAV\text{\_}1k$.

Model	Set	Acmx	Acmx at E	Acmn	Acmn at E	Avacy of 50E
ChestX-ray14	Train	0.9764	47	0.8988	1	0.9688
	Test	0.9780	36	0.9456	3	0.9696
ChexPert	Train	0.9886	30	0.9107	1	0.9810
	Test	0.9862	41	0.5038	1	0.9701
ImageNet	Train	0.8024	44	0.6729	1	0.7925
	Test	0.8259	31	0.7688	1	0.8123
INCXR14	Train	0.9904	34	0.9206	1	0.9823
	Test	0.9828	46	0.5561	2	0.9644
INCxP	Train	0.9583	18	0.8879	1	0.9519
	Test	0.9621	43	0.6965	1	0.9514
iNSup	Train	0.9494	42	0.8666	1	0.9419
	Test	0.9566	24	0.7770	1	0.9467
iNSupFS	Train	0.9594	31	0.8816	1	0.9506
	Test	0.9711	42	0.9284	6	0.9611
iNMSwAV	Train	0.9995	39	0.9363	1	0.9951
	Test	0.9952	23	0.9140	1	0.9892
$MoCo\text{\_}v1$	Train	0.9379	44	0.8179	1	0.9289
	Test	0.9353	45	0.8665	29	0.9256
$MoCo\text{\_}v2$	Train	0.8027	42	0.7447	1	0.7917
	Test	0.8135	47	0.5045	1	0.7961