Table 5.

Shows the class accuracies for each of the eight classes, along with their mean percentile performance and SD.

Lesion	MobileNetV1	MobileNetV2	NASNetMobile	Xception
ak	68.00% ($\pm 13.04$)	80.00% ($\pm 0.00$)	72.00% ($\pm 16.43$)	66.00% ($\pm 20.74$)
vasc	80.50% ($\pm 4.47$)	90.50% ($\pm 2.82$)	78.50% ($\pm 6.52$)	91.00% ($\pm 4.18$)
sk	52.78% ($\pm 7.61$)	67.78% ($\pm 8.97$)	56.11% ($\pm 6.63$)	72.78% ($\pm 10.65$)
df	37.14% ($\pm 12.78$)	68.57% ($\pm 3.67$)	40.00% ($\pm 25.56$)	71.43% ($\pm 14.29$)
bcc	65.11% ($\pm 8.85$)	73.62% ($\pm 10.94$)	61.70% ($\pm 7.82$)	73.19% ($\pm 6.13$)
scc	65.14% ($\pm 2.39$)	89.71% ($\pm 5.75$)	65.14% ($\pm 3.73$)	85.71% ($\pm 4.95$)
mel	85.81% ($\pm 7.77$)	89.03% ($\pm 4.31$)	85.81% ($\pm 7.43$)	87.74% ($\pm 2.70$)
nv	91.38% ($\pm 0.93$)	91.87% ($\pm 2.64$)	92.52% ($\pm 2.47$)	91.00% ($\pm 4.18$)