An Ensemble of Transfer Learning Models for the Prediction of Skin Lesions with Conditional Generative Adversarial Networks

Table 4

Class-wise performance of the ensemble models on balanced and unbalanced datasets.

Ensemble models	Unbalanced dataset			Balanced dataset
Class of skin lesion	Recall (%)	Precision (%)	F1-score (%)	Recall (%)	Precision (%)	F1-score (%)
AKIEC	73.84	85.71	79.33	84.61	94.82	89.43
BCC	80.39	92.13	85.86	90.19	94.84	92.46
BKL	75.34	83.33	79.13	84.93	90.73	87.73
DF	78.26	94.73	85.71	95.65	95.65	95.65
MEL	61.71	84.56	71.35	72.07	92.48	81.01
NV	98.65	91.62	95.00	99.03	93.91	96.40
VASC	96.42	84.37	90	96.42	90	93.10