Ensemble Framework of Deep CNNs for Diabetic Retinopathy Detection

Table 5

Comparison of models.

Models	Imbalance acc.	Balance acc.	Recall	Specificity
Batch_size_2	77.00	60.44	48.36	85.40
Batch_size_3	77.40	61.62	49.87	84.98
Batch_size_4	77.78	60.41	46.33	85.48
Model 1	78.13	60.80	44.85	84.43
DenseNet-121	78.08	—	49.80	85.93
ResNet50	77.28	—	45.14	84.8
Inception-V3	71.25	—	48.17	84.54
Xception	74.41	—	46.26	84.61
Dense169	73.91	—	47.58	84.85
Pratt et al.’s [20] model	75.00	—	30.00	95.00
Model 2	80.36	60.89	47.70	85.94