Table 6.
The comparisons between the customized EyeNet, ResNet50 [40], Inception V3 [44], VGG19 [45], and the proposed E-DenseNet BC with different depths and weights on APTOS 2019 dataset due to ACC, SEN, SPE, DSC, QKS, calculation time (T) in minutes (m) performance measures
| Model | ACC (%) | SEN (%) | SPE (%) | DSC (%) | QKS | T(m) |
|---|---|---|---|---|---|---|
| A customized EyeNet | 75.7 | 76 | 82 | 74.9 | 0.61 | 14 |
| ResNet50 [40] | 67.4 | 70 | 52.6 | 66.2 | 0.51 | 27 |
| Inception V3 [44] | 49.3 | 53 | 51 | 49.1 | 0.13 | 38 |
| VGG19 [45] | 72.5 | 80 | 68 | 71 | 0.55 | 17 |
| E-DenseNet BC-169-ImageNet | 80.2 | 73.6 | 92 | 80 | 0.70 | 5.2 |
| E-DenseNet BC-169 | 80.6 | 72 | 90 | 80.9 | 0.71 | 7 |
| E-DenseNet BC-201-ImageNet | 82.2 | 74.6 | 92.3 | 82 | 0.73 | 10 |
| E-DenseNet BC-121-ImageNet | 72 | 75 | 48.4 | 71.54 | 0.58 | 3 |
| E-DenseNet BC-121 | 84 | 94 | 73 | 83.7 | 0.75 | 4 |