International Journal of Biomedical Imaging

Research Article

Automatic Detection of AMD and DME Retinal Pathologies Using Deep Learning

Table 3

Description of the created CNN and its layers.
BlocksLayersValues
Conv_bloc 1Convolution2D
BatchNormalization2D
Activation Function		Number of filters: 64
Kernel size: (3,3)
Convolution2D-
ReLu
Number of filters: 128
Kernel size: (3,3)
Conv_bloc 2BatchNormalization2D
Activation Function
MaxPooling2D
Convolution2D		-
ReLu
Kernel size: (2,2) with stride: 2
Number of filters: 128
Kernel size: (3,3)
Res_1BatchNormalization2D
Activation Function
Convolution2D		-
ReLu
Number of filters: 128
Kernel size: (3,3)
BatchNormalization2D
Activation Function
Convolution2D		-
ReLu
Number of filters: 256
Kernel size: (3,3)
Conv_bloc 3BatchNormalization2D
Activation Function
MaxPooling2D
Convolution2D		-
ReLu
Kernel size: (2,2) with stride: 2
Number of filters: 512
Kernel size: (3,3)
Conv_bloc 4BatchNormalization2D
Activation Function
MaxPooling2D
Convolution2D		-
ReLu
Kernel size: (2,2) with stride: 2
Number of filters: 512
Kernel size: (3,3)
Res_2BatchNormalization2D
Activation Function
Convolution2D		-
ReLu
Number of filters: 512
Kernel size: (3,3)
MaxPoolBatchNormalization2D
Activation Function
AdaptiveMaxPooling2D
Flatten		-
ReLu
-
-
Bloc_classificationDropout
Dense
Activation Function		0.2
512
Linear