input_sz: Size of the input feature map
kernel_sz = Size of the kernel/filter
stride = Stride length
padding_sz = Amount of zero padding
layer_type = Either 'conv' for standard convolutional layer or 'transposed_conv' for transposed convolutional layer
duration = Duration (in seconds) between gif frames
