Diagnosis of Diabetic Retinopathy through Retinal Fundus Images and 3D Convolutional Neural Networks with Limited Number of Samples

Table 2

Architecture hyperparameters for the proposed 3D-CNN model for multiclass classification task with 10 feature maps in the convolutional layer.

Layer	Filter size	Number of filters	Stride size	Dropout rate	Output size
Conv1+BN+ELU		10	1	—
MaxPool1		—	2	—
Conv2+BN+ELU		10	1	—
MaxPool2		—	2	—
Conv3+BN+ELU		10	1	—
MaxPool3		—	2	—
Conv4+BN+ELU		10	1	—
MaxPool4		—	2	—
Conv5+BN+ELU		10	1	—
MaxPool5		—	2	—
Conv6+BN+ELU		10	1	—
MaxPool6		—	2	—
FC 1	—	500	—	—	500
FC 2	—	300	—	—	300
FC 3	—	5	—	—	5
Dropout	—	—	—	0.1	5
Softmax	—	—	—	—	5

FC: fully connected; MaxPool: maximum pooling; BN: batch normalization; ELU: exponential linear unit; Conv: convolutional.