Figure 6.

Illustration of the proposed 3D-CNN architecture. The network contains three convolutional layers, two max-pooling layers, and two fully connected layers. The number below each layer represents the feature map size after convolution/pooling.