Laplacian needs upsampling of Gaussian filtered output?

[Harimus]

laploss/laploss.py

Line 28 in 7a1025e

t_diff = current - t_gauss

I noticed that the Laplacian on each layer is not conducting up-sampling. If I recall correctly, The equation of Laplacian image at layer i is

In the current implementation it looks like The second term in the equation is not expand(G_{i+1}) but rather, gaussian_smoothing(G_i). Should it not be:

t_gauss_down = downsample(t_gauss) #
t_gauss_up = upsample(t_gauss_down) #Upsampling, could be nearest for example
t_diff = current - t_gauss_up

[Harimus]

Ref: