Classification of Diabetic Retinopathy Severity in Fundus Images Using the Vision Transformer and Residual Attention

Table 8

The effect of the number of attention heads in the DDR dataset. Similar to Table 7, here sensitivity, specificity, accuracy, and AUC are average values, respectively.

Head setting	Sensitivity	Specificity	Accuracy	AUC
Head = 1	0.8037	0.7925	0.7827	0.8436
Head = 4	0.8064	0.8082	0.8241	0.8912
Head = 6	0.7932	0.8030	0.8148	0.8643
Ours (head = 2)	0.8140	0.8245	0.8235	0.9018

The bold font represents the best performance in each column. The italic font represents the second-best performance in each column.