Skip to main content
. 2017 Dec 5;7:16991. doi: 10.1038/s41598-017-17299-w

Table 1.

Architectures of the CNNs for the classification and regression in this study.

classification/regression
Layer Filter size # of filters or elements
Convolution 3 × 3 × 3 16
Leaky ReLU
Batch normalization
Max pooling 2 × 2 × 2
Convolution 3 × 3 × 3 16
Leaky ReLU
Max pooling 2 × 2 × 2
Convolution 3 × 3 × 3 16
Leaky ReLU
Full connected 64/32
Leaky ReLU
Full connected 64/8
Leaky ReLU
Full connected 50/1
Soft max/Linear