Removing Points on corners

[Etienne]

So, I have this...
I want to delete points at the corner. I'm using a sphere, copy to points on my first ADD that have the corner points I need to remove.


I have created  a group name corner points using Bounding Object as Bounding type.
but the group as nothing selected.

Can someone point me out where I might have gone wrong please.
Thanks.

[ThomasPara]

Set your sphere to mesh or polygon.

[Etienne]

@ThomasPara I just did and didnt seem to work...


 

[ThomasPara]

instead you can use a pointwrangle, input your points in the first input, and your cornerpoints in the second.

int nearpt = nearpoint(1,@P,0.0001);
if(nearpt >= 0){
	i@group_corners = 1;
	}

 

[anim]

while even primitive shperes should work, it's difficult to see what's wrong with your scene just from screenshots

alternatively you can use Group Expression (or wrangle)

nearpoint(1, @P, 0.01) != -1

where 0.01 is you threshold which means that only points within this threshold distance from points in second input will be in the group

 

select_points_by_second_input.hip

[Etienne]

@ThomasPara I'm not sure if I fallow you.
Something like this..

 

[Etienne]

Here the file. Will be more easier to see whats wrong.

House_gen.hipnc

[konstantin magnus]

You can identify corner points on flat polygons by their tangents' length between two edges using the polyframe node.

 

corners.hipnc

[Etienne]

I am fallowing this video. at 46 minutes my group just do not work.

House_gen.hipnc

 

 

Edited December 17, 2019 by Etienne

[Sepu]

Not sure why isn't working with the ctp as it is, if you use a transform and move the sphere to the pt it works as well. Anyway, for now if you add a vdbfrompolys after the ctp it works. This might be a bug so it might good to report it. 

 

[ftaswin]

For your case, this works (run on point wrangler):

 

int pts[] = neighbours(0,@ptnum);
if(len(pts)!=2) return;
vector n1 = normalize(point(0,"P",pts[0])-@P);
vector n2 = normalize(point(0,"P",pts[1])-@P);
@group_corners = dot(n1,n2)>-0.1;

 

[Noobini]

methinks there's a bug with the resample ?

simple example, lowres circle, group all points as orig.

resample, now select the orig group....they ALL light up...not kosher I don't think.

 

vu_resample_bug.hiplc

[AntoineSfx]

The spheres / bounding objects is not necessary anyway. Just group the corner points at a point in your workflow when you have just them, then group transfer said group to the refined geo. Then groupcombine to invert the group and you have the notcorners group..

[AntoineSfx]

9 hours ago, Noobini said:

methinks there's a bug with the resample ?

simple example, lowres circle, group all points as orig.

resample, now select the orig group....they ALL light up...not kosher I don't think.

 

vu_resample_bug.hiplc

I didn't find any references in the doc on the subject of how the attributes or group are handled during a resample.

I know the attributes are linearly interpolated, but it's unclear how the groups are interpolated, whether in resample or subdivide.

However, it can be emulated using a float attribute..

FWIW:

the point groups are correctly interpolated in "Treat polygon As: Straight Edges" mode, and not in the two other ones (Subdivision, Interpolate)

Edited December 18, 2019 by AntoineSfx

[Etienne]

@Sepu Thanks alot, using a VDBFromPoly works!
:)
 

[AntoineSfx]

20 hours ago, Etienne said:

I am fallowing this video. at 46 minutes my group just do not work.

House_gen.hipnc

 

 

You have almost duplicate points in Add1 because of  Facet1 (unique points)

The bounding object in group1/ input2 is made of non connected  and overlapping geometries. Each sphere is present twice, and it's causing problems in the algorithm they're using to test if a point is inside a bounding object or not.

See https://en.wikipedia.org/wiki/Point_in_polygon for the reason why two overlapping spheres will cause points at their intersection being incorrectly seen as outside.

The VDB  From Poly is "solving" this by merging the two identical spheres together and makes the geometry nice and clean.

 

Edited December 18, 2019 by AntoineSfx

[Noobini]

7 hours ago, AntoineSfx said:

I didn't find any references in the doc on the subject of how the attributes or group are handled during a resample.

I know the attributes are linearly interpolated, but it's unclear how the groups are interpolated, whether in resample or subdivide.

However, it can be emulated using a float attribute..

FWIW:

the point groups are correctly interpolated in "Treat polygon As: Straight Edges" mode, and not in the two other ones (Subdivision, Interpolate)

well Treat polygon as straight edges is STILL wrong for me. Anyway, to me it's far more logical to NOT auto lump any new geo into existing groups...just me.

[Noobini]