Skip to main content

Table 1 The averaged accuracy of DR grading among different methods

From: Development of revised ResNet-50 for diabetic retinopathy detection

 

Multiply

Add

Average

Maximum

Subtract

Training accuracy

0.7876

0.7892

0.8377

0.82342

0.88048

Training loss

0.5747

0.57436

0.5533

0.56892

0.51524

Validation accuracy

0.74326

0.74696

0.73142

0.74088

0.7311

Validation loss

0.6077

0.5979

0.615

0.61406

0.6376

Test accuracy

0.77224

0.76076

0.76076

0.7628

0.75064

  1. The best performance is highlighted in bold