Closed Hansello218 closed 7 months ago
it is a parametric trick (logrithim of s), now trbfoutput is in range (0 to1) to make sure final opacity range is (0 to1).
if you take the rbf function return torch.exp(-1*x.pow(2))
, you can get
torch.exp(-1*(offset / exp(trbfscale)).pow(2))
, same as
torch.exp(-1*(offset * exp(-trbfscale)).pow(2))
, where
-1*(offset * exp(-trbfscale)).pow(2)
is same as -1*offset.pow(2)* exp(-trbfscale).pow(2)
,
is same as
-1* exp(-trbfscale).pow(2)*offset.pow(2)
-1* exp(-2trbfscale)*offset.pow(2)
It's correct.Thanks a lot!
When using radial basis functions to represent opacity, I noticed that the computation method provided in your code does not match the calculation method given in the equation 7 of the paper. Could you please clarify this?