. 2021 Jan 19;11:1839. doi: 10.1038/s41598-021-81525-9

Table 2.

Test results of the nine models with different depth levels as well as three baseline models.

Base network	Depth^a	F1 score			AUC^b	Accuracy	Soft accuracy
Base network	Depth^a	OAP	IN	OBS	AUC^b	Accuracy	Soft accuracy
MobileNetV1 (a = 0.25)	0	0.56	0.90	0.73	0.97	0.85	0.95
MobileNetV1 (a = 0.25)	1	0.64	0.90	0.74	0.97	0.85	0.95
MobileNetV2 (a = 0.35)	0	0.56	0.78	0.46	0.86	0.67	0.76
MobileNetV2 (a = 0.35)	1	0.77	0.81	0.52	0.89	0.73	0.82
VGG16	0	0.78	0.91	0.76	0.97	0.86	0.94
VGG16	1	0.81	0.93	0.80	0.98	0.90	0.97
InceptionV3	0	0.38	0.88	0.68	0.95	0.81	0.92
InceptionV3	1	0.36	0.87	0.67	0.95	0.81	0.92
ResNet50V2	0	0.62	0.90	0.70	0.96	0.84	0.94
ResNet50V2	1	0.71	0.90	0.72	0.97	0.85	0.94
InceptionResNetV2	0	0.66	0.88	0.65	0.96	0.83	0.93
InceptionResNetV2	1	0.78	0.92	0.79	0.98	0.89	0.96
DenseNet121	0	0.67	0.90	0.73	0.97	0.85	0.93
DenseNet121	1	0.71	0.90	0.74	0.97	0.86	0.93
NASNet Mobile	0	0.57	0.78	0.58	0.88	0.70	0.80
NASNet Mobile	1	0.49	0.71	0.63	0.84	0.65	0.76
Xception	0	0.51	0.88	0.65	0.95	0.82	0.93
Xception	1	0.63	0.89	0.69	0.96	0.84	0.93
CBR-LargeT	–	0.69	0.90	0.77	0.96	0.85	0.93
CBR-Small	–	0.75	0.91	0.72	0.97	0.86	0.95
CBR-Tiny	–	0.61	0.84	0.74	0.92	0.78	0.88

Bold text indicates the highest value among the models.

CBR convolution, batch-normalization, ReLu-activation²⁹.

^aThe definition of depth is provided in the Supplementary Material.

^bMicro-averaged AUC score.