Skin Cancer Segmentation and Classification Using Vision Transformer for Automatic Analysis in Dermatoscopy-Based Noninvasive Digital System

Table 6

Classification report of the best combination for each transfer learning model.

Model		Precision	Recall	-score	Model		Precision	Recall	-score
ViT (Google)	Benign	0.96950	0.95300	0.96120	ViT (VAN)	Benign	0.93620	0.91000	0.92290
	Malignant	0.95380	0.97000	0.96180		Malignant	0.91250	0.93800	0.92500
	Accuracy	0.96150				Accuracy	0.92400
	Macro-	0.96150				Macro-	0.92400
	Weighted-	0.96150				Weighted-	0.92400
ViT (Mae)	Benign	0.94230	0.89800	0.91960	ViT (BEiT)	Benign	0.91430	0.80000	0.85330
	Malignant	0.90260	0.94500	0.92330		Malignant	0.82220	0.92500	0.87060
	Accuracy	0.92150				Accuracy	0.86250
	Macro-	0.92150				Macro-	0.86200
	Weighted-	0.92150				Weighted-	0.86200
ViT (ResNet50)	Benign	0.81470	0.80000	0.80730	ViT (DiT)	Benign	0.83160	0.80000	0.81550
	Malignant	0.80350	0.81800	0.81070		Malignant	0.80730	0.83800	0.82240
	Accuracy	0.80900				Accuracy	0.81900
	Macro-	0.80900				Macro-	0.81890
	Weighted-	0.80900				Weighted-	0.81890