An example architecture of the 2D convolutional layer. The light blue cube represents the input feature map, the blue small cube denotes the filter, and the light orange cube represents the output feature maps. Each neuron in the output layer is obtained by multiplying the corresponding elements of the input feature maps and filter.