Detecting representative characteristics of different genders using intraoral photographs: a deep learning model with interpretation of gradient-weighted class activation mapping

Table 2 Performance of the R-net model for female detection in the testing set

Group	Actual Female		Actual Male		Sensitivity	Specificity
Group	True Positive	False Negative	True Negative	False Positive	Sensitivity	Specificity
Uncropped Children Group	28	19	44	2	0.596	0.933
Cropped Children Group	29	18	42	4	0.617	0.879
Uncropped Youth Group	49	5	52	1	0.907	0.98
Cropped Youth Group	44	10	50	3	0.815	0.936
Uncropped Testing Samples	77	24	96	3	0.762	0.970
Cropped Testing Samples	73	28	92	7	0.723	0.929

ISSN: 1472-6831