[Soc-2017-dev] Weekly Report #07 - Silhouette Operator
Sebastian Witt
W.Basti at gmx.de
Tue Jul 18 10:21:54 CEST 2017
Hi Sergey,
you are right, it is not finished. What i was trying to say is that it
doesn't need any new features or methods to generate the geometry.
It is just the methods present need some bugfixes.
Bugs i can currently identify:
1. T-intersection center quads are Tris (holes in the geometry)
2. Intersections disappear/ are placed in infinity (division by 0 or
non-initialized memory access)
3. T-intersection rotation out of order
The normal orientation should be an easy fix because the algorithms
already support flipping and calculating the orientation of a polygon is
a common problem.
I hope this clarifies what i was trying to say,
Sebastian Witt
On 16.07.17 13:29, Sergey Sharybin wrote:
> Hi,
>
> When you mention that used one method which is better than other one,
> it is always a good idea to show two pictures: one with old method and
> one with new one. This helps quantifying how much better results from
> user perspective.
>
> What i didn't understand is: you're mentioning that geometry is fully
> generated and algorithm is finished, but in the next sentence you say
> that there are lots of issues. So either something is not finished or
> there shouldn't be issues.
>
> Is the normal orientation the only problem now?
>
> On Thu, Jul 13, 2017 at 7:54 PM, Sebastian Witt <W.Basti at gmx.de
> <mailto:W.Basti at gmx.de>> wrote:
>
> Weekly Report #07 - Silhouette Operator
>
> I am sending my weekly report a little earlier this week because I
> am heading for an earlier weekend so i will be missing the friday.
> I tried to do some extra hours on monday and tuesday to get some
> extra progress and will do so in the next week as well.
>
> Nevertheless i made some good progress this week. Last week I
> decided to rework the geometry generation of the silhouette operator.
>
> The unified interpolation Method
> The major advantage of the new system is a unified interpolation
> method.
> The method decides the shape of the object in the 3rd Axis which
> is not defined by the sculptor. Therefore i can easily control the
> whole output by altering this method.
> Currently it controls the smoothness and depth of the generated
> geometry. The smoothness is a value between 0 and 1.
> With 0 being an object with sharp edges. If you draw a circle for
> example this result in a cylinder.
> A smoothness value of 1 will smooth out the corner and results in
> a sphere.
>
> Order of Generation
> Before the rework the different parts were relying on each other.
> In other words, i generated tube-parts first carried bordering
> edges/vertices over and generated T-intersections as well as
> Cap-parts afterwards. Due to this dependency i had to incorporate
> a lot of corner cases and experienced a lot of unexpected
> behaviour. With the new system every part is self-contained and
> exposes only one continuous edgeloop. This makes it easier to
> track done errors and is a cleaner approach in general.
>
> Connecting the Parts
> With those continuous edgeloops it is easier now to connect the
> different parts. There is still some issues with the general
> orientation of some parts and inverted edgeloops. But that should
> not be a big problem to solve.
>
> What is the current stage?
> In theory the geometry is now completely generated. The algorithm
> itself is done! In practice there are still a lot of issues with
> the system. As mentioned above the part orientation is messed up
> sometimes resulting in errors on the connections.
>
> What is up next?
> Next step is to fix the issues and make the tool a little more
> stable. After that i will try to fix the normal orientation.
> Currently if drawn counterclockwise it is correctly oriented and
> if drawn clockwise the normals need to be flipped.
> If the silhouette is more stable I will start exposing the tool
> settings and hopefully move on soon to connecting to existing
> geometry.
>
> Best regards,
>
> Sebastian Witt
>
>
> _______________________________________________
> Soc-2017-dev mailing list
> Soc-2017-dev at blender.org <mailto:Soc-2017-dev at blender.org>
> https://lists.blender.org/mailman/listinfo/soc-2017-dev
> <https://lists.blender.org/mailman/listinfo/soc-2017-dev>
>
>
>
>
> --
> With best regards, Sergey Sharybin
>
>
> _______________________________________________
> Soc-2017-dev mailing list
> Soc-2017-dev at blender.org
> https://lists.blender.org/mailman/listinfo/soc-2017-dev
-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://lists.blender.org/pipermail/soc-2017-dev/attachments/20170718/77ffc75a/attachment-0001.htm
More information about the Soc-2017-dev
mailing list