Fig. 3.
Overview of the 3D CNN structure. Input map and output of each feature extraction layer are visualized. C1 represents convolutional layer 1, which contains 32 types of kernels or filters of size Ȉ with a stride step 2; C2 represents convolutional layer 2, which contains 32 types of kernels of filters of size 3 with a stride step 2;C2 represents pooling layer with pooling kernel size of 2; FC represents fully connected layer, with 128 nodes in this layer; Output layer contains 10 nodes representing each class of the 10 RSNs labels.