Skip to main content
. 2020 Aug 1;20(15):4301. doi: 10.3390/s20154301

Table 3.

Some key parameters in the nodule candidate detection sub-network (m represents the channels of the former layer)

Modules Parameters
U-Net Conv1 3×3×64, stride=1
Conv2 3×3×128, stride=2
Conv3 3×3×256, stride=2
Conv4 3×3×512, stride=2
Deconv1 3×3×512, stride=2
Deconv2 3×3×256, stride=2
Deconv3 3×3×128, stride=2
Deconv4 3×3×64, stride=2
Inception Conv1 1×1×(m/2), stride=1
Conv2 1×1×(m/2), stride=1
Conv3 1×3×(m/2), stride=1
Conv4 3×1×(m/2), stride=1
Conv5 1×1×m, stride=1
Dilated convolution Di-conv1 3×3×m, stride=1, dilationrate=2