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

Table 6

Performance comparison of different models on the DDR dataset.

Model	0	1	2	3	4	5	AA
VGG-16 [64]	0.9537	0.0423	0.5625	0.3944	0.6436	0.9422	0.5898
ResNet-18 [65]	0.9548	0.0582	0.622	0.3662	0.5818	0.9133	0.5827
GoogLeNet [66]	0.9574	0.0265	0.5759	0.3380	0.5782	0.9162	0.5654
DenseNet-121 [67]	0.8930	0.2275	0.5751	0.4085	0.6364	0.9306	0.6119
SE-BN-Inception [68]	0.9452	0.0476	0.6458	0.1268	0.5818	0.9046	0.5418
ViT + CSRA	0.8280	0.9635	0.7793	0.9881	0.9581	0.9752	0.9154

The bold font represents the best performance on that class. The italic font indicates the second-best performance on that class.