Author Topic: sculpting tools  (Read 28174 times)

RMS

  • meshmixer founder
  • Administrator
  • Hero Member
  • *****
  • Posts: 1238
    • View Profile
    • gradientspace
Re: sculpting tools
« Reply #45 on: May 21, 2011, 11:01:33 PM »
Yea, I have looked at this stuff, but for dynamic meshing it is not so useful. The problem is that they are generally mean for processing large blocks of data in batch. But when the mesh is changing each frame, it is probably going to actually be *more* expensive to use the GPU. I think MudBox uses CUDA, but it has a static mesh structure (when you are sculpting)



created meshmixer - now starting gradientspace - meshmixer consulting available http://www.gradientspace.com/consulting

HolyEnigma74

  • Global Moderator
  • Full Member
  • *****
  • Posts: 246
    • View Profile
Re: sculpting tools
« Reply #46 on: May 21, 2011, 11:31:13 PM »
oh i kinda see what you're saying..
3D-Coat Does it With Voxels so i figured it could be done

RMS

  • meshmixer founder
  • Administrator
  • Hero Member
  • *****
  • Posts: 1238
    • View Profile
    • gradientspace
Re: sculpting tools
« Reply #47 on: May 21, 2011, 11:34:18 PM »
ah, well a voxel grid is just a 3D texture. It has fixed connectivity (each cell has 26 nbrs).

In a dynamic mesh, each vertex has K neighbours, and those neighbours change. So it is basically a large dynamic graph. Which, unfortunately, is the kind of problem that GPU-style stream processing is really bad at handling. This is why OpenGL and Direct3D don't support a "mesh" datatype - you have to break it down into a buffer of triangles or quads (if you have done graphics programming...)
created meshmixer - now starting gradientspace - meshmixer consulting available http://www.gradientspace.com/consulting

andreicirdu

  • Newbie
  • *
  • Posts: 29
    • View Profile
    • 3d concepts
Re: sculpting tools
« Reply #48 on: May 23, 2011, 09:24:50 AM »
The mesh seems to be tesselated under the whole diameter of the brush, even where the surface isn't visibly raised by the brush. Is it really necesary to have that tessellation? Can't it be optimized to tesselate only where the surface experiences visible height/depth changes? There should be a cutoff to the tesselation where the surface isn't visibly changed...only if it offers an advantage over the current impementation.

iatriki.carbonmade.com

RMS

  • meshmixer founder
  • Administrator
  • Hero Member
  • *****
  • Posts: 1238
    • View Profile
    • gradientspace
Re: sculpting tools
« Reply #49 on: May 23, 2011, 10:38:06 AM »
yup, I agree entirely. I haven't got that far, though, and doing that is quite a bit more complicated.

Basically there are various different ways to do things:
1) uniform refinement of brush area
2) refinement of brush area taking brush mask into account
3) refinement of brush area taking deformation into account

I am at (2) right now. Doing (3) is the next step, and the non-uniform reduce brush uses my initial implementation. But it is not robust enough yet for brushing. The problem for brushing is that to get nice-looking strokes, you want the uniform refinement. So when you go to adaptive refinement, the obvious approaches work like 95% of the time, but that other 5% shows up several times in any stroke, as ugly artifacts.

So, it's getting there...
created meshmixer - now starting gradientspace - meshmixer consulting available http://www.gradientspace.com/consulting