Contrast Enhancement in Mammograms Using Convolution Neural Networks for Edge Computing Systems

Table 1

Layers size.

Layer	Kernel size	Input size	Output size
Convolution layer 1	5 × 5	3 × 256 × 256	64 × 256 × 256
Convolution layer 2	3 × 3	64 × 256 × 256	64 × 256 × 256
Upsampling layer 1	—	64 × 256 × 256	64 × 512 × 512
Convolution layer 3	3 × 3	64 × 512 × 512	64 × 512 × 512
Max-pooling layer	2 × 2	64 × 512 × 512	64 × 256 × 256
Convolution layer 4	3 × 3	64 × 256 × 256	64 × 256 × 256
Upsampling layer 2	—	64 × 256 × 256	64 × 512 × 512
Convolution layer 5	3 × 3	64 × 512 × 512	64 × 512 × 512
Max-pooling layer	2 × 2	64 × 512 × 512	64 × 256 × 256
Convolution layer 6	3 × 3	64 × 256 × 256	64 × 256 × 256
Convolution layer 7	3 × 3	64 × 256 × 256	64 × 256 × 256
Convolution layer 8	3 × 3	64 × 256 × 256	64 × 256 × 256
Fully connected layer	—	64 × 256 × 256	3 × 256 × 256