Table 6.

Comparison evaluation performances of ROC and PR curves for backbone and proposed models using SGD optimizer.

	Learning Rate/Loss Function	Macro-Average	Micro-Average	COVID-19	Lung Opacity	Normal	Pneumonia
(1) ROC Evaluation Curve
Backbone	${10}^{-4}$/Categorical smooth loss	0.92	0.92	0.87	0.96	0.91	0.92
Proposed		0.92	0.92	0.86	0.96	0.93	0.93
Backbone	${10}^{-3}$/Categorical smooth loss	0.92	0.92	0.96	0.97	0.89	0.86
Proposed		0.96	0.96	0.99	0.96	0.95	0.95
Backbone	${10}^{-4}$/Categorical
cross-entropy	0.92	0.92	0.90	0.92	0.86	0.92
Proposed		0.93	0.93	0.92	0.97	0.93	0.89
Backbone	${10}^{-3}$/Categorical cross-entropy	0.60	0.60	0.63	0.53	0.47	0.77
Proposed		0.93	0.93	0.88	0.96	0.94	0.93
(2) Precision-Recall (AP) Evaluation Curve
Backbone	${10}^{-4}$/Categorical smooth loss	0.79	0.77	0.88	0.70	0.88
Proposed		0.80	0.75	0.82	0.77	0.90
Backbone	${10}^{-3}$/Categorical smooth loss	0.80	0.92	0.88	0.67	0.80
Proposed		0.90	0.93	0.91	0.85	0.92
Backbone	${10}^{-4}$/Categorical cross-entropy	0.81	0.81	0.86	0.74	0.88
Proposed		0.82	0.83	0.93	0.74	0.84
Backbone	${10}^{-3}$/Categorical cross-entropy	0.31	0.31	0.29	0.25	0.65
Proposed		0.82	0.82	0.86	0.76	0.90