A grayscale image (a single channel) is about to go through one of two first layers. Both must hand the next stage the same number of values.
Design C (convolutional). filters, each . Each filter slides across the image one step at a time, stops only where its whole window fits inside the image, and produces one feature map.
Design F (flat). The image is unrolled into a list and fed to an ordinary layer where every input connects to every hidden unit. To keep the comparison fair, this layer gets exactly as many hidden units as Design C outputs values in total.
Count weights only and ignore biases.
How many times more weights does Design F hold than Design C?
Give the exact integer.