Skip to main content
. 2022 Apr 26;14:871706. doi: 10.3389/fnagi.2022.871706

Table 3.

Detailed parameters of our 3D-CNN subject sub-network.

Layer name Kernel Stride Activation Output channels
Conv3D 3 × 3 × 3 1 BachNorm3d+Relu 32
MaxPooling3D 3 × 3 × 3 2 32
Conv3D 3 × 3 × 3 1 BachNorm3d+Relu 128
MaxPooling3D 3 × 3 × 3 2 128
Conv3D 3 × 3 × 3 1 BachNorm3d+Relu 256
MaxPooling3D 3 × 3 × 3 2 256
Conv3D 2 × 2 × 2 2 BachNorm3d+Relu 512
MaxPooling3D 5 × 5 × 5 2 512
Globel-Avg-Pooling3D 1 × 1 × 1 512