> If it is easy, you might suggest the code correction (on beta code ?).

It's very easy.


> The paper looks like scholar... but fails to have an abstract, explicit
> demonstration and rely a lot on asserting evidences (without formal
> demonstration/mesurement/estimation).

The paper gives a mathematical derivation for the formulation.
Furthermore, omiting the initial projection of the gradient before subtraction
(from the original normal) is not only theoretically wrong but as shown in
figure 2 also visually wrong.

Cheers

Post a reply to this message