In Development: Subdivide and Smooth

[justry]

Yeah I know a second iteration wasn't necessary, I just wanted to test the performance of the script on a high-poly model.

The first one took you 15 seconds though??! My machine's gotta be real slow . Gotta get me a new one!

Thanks for the quick reply Rick, and your script's great too man. All of them actually .

And waiting for that amazing script of yours whaat!

[RickW]

I think our machines are comparable processor speed, so maybe it's a memory issue (running 4GB RAM here). Regardless, my script is slow compared to Whaat's, since I didn't really get past the proof-of-concept stage into optimizing.

[Whaat]

Subdivide and Smooth version 1.0 Final Features List

• Low-poly proxy subdivision modeling
• Automatic updating of subdivision surfaces while editing a proxy (optional)
• Material inheritance of proxy model to subdivision surface (optional)
• Creasing tool - allows toggling of fully creased edges or vertices
• Up to 4 subdivision and smooth iterations
• Unlimited proxy subdivision models can exist in the same file
• Subdivision and smoothing of a single isolated face
• Robust subdivision algorithm that can handle almost any geometry
• Subdivision without smoothing (linear subdivision) - also triangulates all meshes and maintains material assignment
• Mesh smoothing (vertex averaging) without subdivision - maintains material assignment
• Extrude tool - useful for modeling clean geometry that subdivides well
• Knife subdivide tool (aka 'Zorro')

Testing begins this week! (sorry, invitation only). I will keep you all posted on the release date!

Whaat

[remus]

Mmmmmm, exciting stuff, i look forward to it whaat, thanks again for all the hard work

[kwistenbiebel]

I am not sure what all the features exactly mean, but it does sound really cool .

[a4chitect]

whaat you rock! I can't wait for this release! you are one of those for whome the human cloning was meant in the first place its interesting that you mention mesh smoothing (without subdivision), does it mean we will be able to smooth contour generated surfaces so they will be more natural in shape? good luck with testing

[Fletch]

can't wait to see it, Whaat...
(on the 'invite' list, on the 'invite' list, on the 'invite' list)

[solo]

By the look and feel of it I think the name should be 'Organix' as it does give one the abllity to model organically.

[Whaat]

@a4chitect said:

does it mean we will be able to smooth contour generated surfaces so they will be more natural in shape? good luck with testing

Can someone post a complicated contour generated surface? I would like to test this. Thanks.

[remus]

Is this any good whaat? might be a bit simple for your purposes, but it was good fun to make

---

whaat contours.skp

[Whaat]

@remus said:

Is this any good whaat? might be a bit simple for your purposes, but it was good fun to make

Thanks remus,

Here are the results:

First image is the original surface generated from contours
Second image is the same surface after performing vertex averaging (same number of polys)
Third image is the surface after performing two subdivision and smooth iterations

It seems like vertex averaging will be very useful to generate more realistic landscape surfaces without increasing poly-count. The only problem is that the the resulting surface will no longer coincide with the contours, so real-world accuracy is lost. There is no easy way to change this so don't ask

---

---

---

[todd burch]

@whaat said:

There is no easy way to change this so don't ask

Just tell them it's an erosion feature thrown in for no charge.

[Mike Lucey]

@unknownuser said:

@whaat said:

There is no easy way to change this so don't ask

Just tell them it's an erosion feature thrown in for no charge.

[a4chitect]

brilliant whaat - exactly as expected - the surface is much more natural, as for the accuracy - I guess the the vertex-averaged surface could meet the controur lines by simply moving it along the Z axis a bit, and even if they can't touch everywhere - the result is better for its smoothness, you can't take the contours for benchmarking since their accuracy is fairly limited too

maybe performing a test could prove my point>

1. create a curved surface with the Smoove tool

2. generate contours either by the dedicated script, or by intersecting with equidistant planes

3. group contour lines and create a surface by using the sandbox tools

4. paint this surface a different color from the original

5. make a copy (change the color) of this new surface and apply your script

6. compare original surface with both new surfaces

whaat thanks for making sketchup better, can't wait for you to release.

[GUNAWAN W]

oh dear God, please give this guy more power to made this script available to everyone soon.

[kwistenbiebel]

Yummy !
Terrific results as far as I see.

This together with a 'wrap texture around object' feature could possibly make other 3d party organic model software trash can material.

Good stuff Whaat!

[wikii]

I wrote script below to simulate whaat's.It works,but not perfect

class Smooth_contours
   attr_reader ;author,;date,;ver
   def initialize
      @author="Wikii"
      @date="20080402"
      @ver="0.01"
      @mm=Sketchup.active_model
      @fs=@mm.selection.to_a.delete_if{|x| x.class!=Sketchup;;Face}
      do_Smooth_contours if @fs.length>0
   end
   def center_point point3ds
      re=Geom;;Point3d.new(0,0,0)
      num=0
      0.upto( point3ds.length-1){|x|
         if point3ds[x].class==Geom;;Point3d
            re.x+=point3ds[x].x
            re.y+=point3ds[x].y
            re.z+=point3ds[x].z
            num+=1.0
         else #weighted
            re.x+=point3ds[x][0].x*point3ds[x][1]
            re.y+=point3ds[x][0].y*point3ds[x][1]
            re.z+=point3ds[x][0].z*point3ds[x][1]
            num+=point3ds[x][1]
         end
      }
      re.x/=num
      re.y/=num
      re.z/=num
      re
   end
   def open_vertex? v
      v.edges.each{|x|
         return true if x.faces.to_a.delete_if{|y| !@fs.include?(y)}.length<2
      }
      return nil
   end
   def face_center fc
      Geom;;Point3d.new center_point(fc.vertices.to_a.collect{|xx| xx.position})
   end
   def do_Smooth_contours
      group_entities=@mm.active_entities.add_group.entities
      vertices=[]
      vertices_info={}
      @mm.start_operation "123"
      n=0
      @fs.each{|x|
         n+=1.0
         Sketchup;;set_status_text "Finished #{(n/@fs.length*1000).to_i/10.0}%"
         cen=face_center(x)
         x.vertices.each{|y|
            if !vertices.include?(y) and !open_vertex?(y)
               vertices<<y
               #~ tmp0=y.position
               #~ tmp=center_point(y.faces.to_a.collect{|z| [face_center(z),z.area]})
               tmp=center_point(y.faces.to_a.collect{|z| face_center(z)})
               #~ tmp0.z=(tmp0.z+tmp.z)/2.0
               #~ tmp0.z=(tmp0.z+tmp.z)/2.0
               #~ vertices_info[y]=center_point([tmp,tmp0]           )
               vertices_info[y]=tmp
            end
         }
      }
      @ee=Sketchup.active_model.active_entities
      puts @ee
      vertices.each{|x|
         t=Geom;;Transformation.new(vertices_info[x]-x.position)
         @ee.transform_entities(t, x)
      }
      @mm.commit_operation
   end
end
Smooth_contours.new

[a4chitect]

thanks wikii, nice script you've put together so quickly! here is a resulting surface for illustration. and your script with a plugins menu entry added to just place in the plugins folder

---

wikii's smooth surface script

---

illustration comparing surface accuracy after smooth

[RayOchoa]

It would be great if you could do a script that could make high poly models low poly. It would be awsome.

[jenujacob]

holy moly! this is simply terrrific!
how long did the sub division take coen?