Table 9.
The comparisons between the customized EyeNet, ResNet50 [40], Inception V3 [44], VGG-19 [45], and the proposed E-DenseNet BC with different depths and weights on IDRiD dataset due to ACC, SEN, SPE, DSC, QKS, and and calculation time (T) in minutes (m) performance measures
| Model | ACC (%) | SEN (%) | SPE (%) | DSC (%) | QKS | T(m) |
|---|---|---|---|---|---|---|
| A customized EyeNet | 45 | 63 | 35 | 44.5 | 0.24 | 17.05 |
| ResNet50 [40] | 32.5 | 38 | 0 | 32.5 | 0 | 23.5 |
| Inception V3 [44] | 32.5 | 40 | 22 | 32.5 | 0 | 27.5 |
| VGG19 [45] | 33 | 40 | 22 | 32.5 | 0 | 16.4 |
| E-DenseNet BC-169-ImageNet | 66.3 | 70 | 49 | 66 | 0.53 | 6 |
| E-DenseNet BC-169 | 61.4 | 70 | 43 | 60 | 0.46 | 4 |
| E-DenseNet BC-121-ImageNet | 64.2 | 61.3 | 50 | 63.8 | 0.49 | 3 |
| E-DenseNet BC-201-ImageNet | 62.2 | 61 | 55 | 61.1 | 0.48 | 7 |
| E-DenseNet BC-121 | 93 | 96.7 | 72 | 96 | 0.94 | 3 |