[Bf-committers] GSoC 2011: Improving Retopology Tools

[Tom M]

On Sun, Apr 3, 2011 at 1:22 PM, Sergey Kurdakov <sergey.forum at gmail.com> wrote:
> Hi Tom
>
>>No, quad dominant remeshish is much more specialized and uses
>>algorithms wholly inappropriate for what this student is working on.
>
> so rohith was intended to use pen to outline simplified area, after he
> finished integration into blender,
> and it is inappropriate - how do you know? because your first reaction
> was it is wrong and you will 'press ahead'

I managed last years Google Summer of Code for blender, helped design
many of the students proposals (including helped rohith), etc.

I also am managing this years GSoC including adding the retopo proposals :)

> the point here - to use algorithm - and it is already integrated with
> Blender and can be further used ( or not ).

Rohiths code can only work on extremely simple meshes because the math
library he needed for accelerating the process was not integrated yet
(it is something he plans to do, but he isn't done yet), there are
other issues as well that make it not production ready.  Even if it
were production ready, the method it uses for generating the new mesh
is totally inappropriate for the simple nature of the retopology tools
that are for GSoC.


>
> why to use? because the algorithm is much more powerfull and can
> retopo quite difficult areas unlike script you promote.

The script I suggested was so that the student could find where in the
code to look for using the apis she will need to know.  She can use
the APIs in the referenced script directly if she is comfortable with
python, so search the source code to find the C API that the python
code in the script is calling.

It was not an alogithmic reference :)


> of cause - the trade of - how difficult is to use, but your point that
> it is inappriate is overstretched - the algo can be used to
> remesh with given quality sketched mesh..

None of your suggested references provide algorithms for generating
the mesh in an appropriate manner.  They are either drastically too
complicated (involving things like taking derivatives of curvature and
other fun math) or way to simple (doing scan fills that result in tris
or quads that will give ugly topology).

LetterRip