Rendering Height Maps In Maya Api Without Using Opengl

[katisss]

hi,
i have to rendering height maps in maya api without using openGL (as farm machines might not have it)
for a constraint. So i am thinking what would be the best aproach to do that?

[rob-bateman]

displacementShader?

[joojaa]

QUOTE(katisss @ 06/17/09, 03:27 PM) [snapback]309668[/snapback]
So i am thinking what would be the best aproach to do that?

Displacement on plane.

[katisss]

wow that was quick. There happen to be any examples in devkit?
EDIT: there is one, just having a look
I have a custom node that is the constraint and it needs the hightmap as an input.
The terrain already exists as a model.
Did i ever say thank you for answering all my questions in no time? Has been a great help.

[joojaa]

You dont need a node its allready implemented as part fo the shading networks. Just surf to the planes sg and click in the dipslacement field and assign the texture a to displacement. Its also WAY better then what a height field is. For example this is just a nurbs sphere with a displacement shader slapped on it:

PS: yes that means you could just use nurbs spheres and displace them into shapes of cars. Its been done by the way, for fun you know.

[katisss]

The geo constraint already works by creating hightmaps to find where stuff should be constrained to.
Now they dont like that this is doen using opengl and wont work on farm machines.
We dont use displacement at all or am i not getting something?

[joojaa]

In any case for your constraint question yes. You also dont need a node for that maya also implements all you need for that use closest point on node to dind the closest uv than input that to placement of file node and read input form that

[joojaa]

QUOTE(katisss @ 06/17/09, 04:28 PM) [snapback]309677[/snapback]
The geo constraint already works by creating hightmaps to find where stuff should be constrained to.

Well in that case just bake the result on export to farm. This si the sort fo step you should have in place in anycase not just for this thing. It saves you from a lot of monkeywrench debugging.

QUOTE(katisss @ 06/17/09, 04:28 PM) [snapback]309677[/snapback]
We dont use displacement at all or am i not getting something?

You probably should

QUOTE
Now they dont like that this is doen using opengl and wont work on farm machines.

The constraint is not actually getting the data form opengl, at least its very unlikely. just dont work that way. Its porbably just reading the raw data directly.

[katisss]

That sounds interesting, i will have to have a closer look at the closest point on mesh thing.
If i badly wanted the heightmaps anyways... how would i go?

The maybe easier question is to know if a linux machine has Opengl available.

[joojaa]

QUOTE(katisss @ 06/17/09, 04:39 PM) [snapback]309680[/snapback]
If i badly wanted the heightmaps anyways... how would i go?

Create diplacement. its by fer the most effective and fastest way to go about. It is in fact faster than a heightfield.

[katisss]

But it is not the workflow used.
Modells already exist and there is not displacement.
I was going to check if MRenderUtils can help me getting the map.
Its an existing project so the decision is as usually not all mine.
The problem is similar to this http://www.highend3d.com/boards/index.php?...amp;mode=linear except that i know what mesh or meshes i am hitting.
I should have called the topic: create heightmap from existing mesh without using opengl.

We only use the height map to look up values, if we can get the values
without the map there should be no need for that at all.

MRenderUtil::raytraceFirstGeometryIntersections, MfnMesh intersect and MMeshIntersector
ook the most promising so far. Anybody has an idea how thy compare speedwise?

[petershipkov]

You are not clear about what exaclty you want to do with this "offset".
But here are two ideas:
Take a look at the "Toon" tools. They can create "live offset" mesh around your geometry based on some texture (bitmap or procedural).
Do this: create a sphere, Toon-AssignOutline, set ProfileLines to offsetMesh. Create new (or connect existing) texture as lineOffsetMap, set lineOffset to something different than 0. Turn-off the drawingOverrides of the newly created mesh. Done.

You can also do this programatically very easy:
Checke the MDynamicsUtil class - it has only one method that returns RGBA pixel values of provided texture (once again - bitmap or one of the built-in procedural texture nodes) based on two input arrays that contain U and V values.

[joojaa]

QUOTE(katisss @ 06/17/09, 05:04 PM) [snapback]309683[/snapback]
I should have called the topic: create heightmap from existing mesh without using opengl.

Oh in that case juts render a image form top and take out the zbuffer. Rednerers can do this with tehuir shadowmap functions. TJis is also wa faster than trying to sample it yourself.

QUOTE
Modells already exist and there is not displacement.

So whats the problem? If its normal gemmetry then opengl never comes into play. The geometry constraint is NOT working on opengl data.

QUOTE
Its an existing project so the decision is as usually not all mine.

Your implementing it so ALL THE DESIGN DECISSIONS are yours no matter what others in the company say. Ultimately its your head. So since that is the case its your choice 100% of the way. What you do After the current step offcourse you want to reuse as much as possible but tahts just common sense. If anybody isnist otherwise then GET THAT IN WRITING!

A few observations tough:

You are doing something for the render slaves It is as a general rule not sane to make computation nodes for render slaves. This all just increases the risk of failure. Its far better to cache whatever you need thats a pretender doen once before submit of before each frame IS a better more easily fixable solution.. It also takes less than 10 minutes of dev job. After all theres plenty of them to do stuff AT submission stage. So as a general rule you should allwatys bake down as much of the animation to simple look ups before sent to frarm in any case. Thsi amkes renders faster and more stable

[katisss]

QUOTE
Oh in that case juts render a image form top and take out the zbuffer. Rednerers can do this with tehuir shadowmap functions. TJis is also wa faster than trying to sample it yourself.

Thats what i want to do but i dont get how. Alll the examples are hardware stuff.
How would i use the software renderer through the api?

QUOTE
So whats the problem? If its normal gemmetry then opengl never comes into play. The geometry constraint is NOT working on opengl data.

It uses a hardware cg shader to do the render and lookup now.

QUOTE
You are doing something for the render slaves It is as a general rule not sane to make computation nodes for render slaves. This all just increases the risk of failure. Its far better to cache whatever you need thats a pretender doen once before submit of before each frame IS a better more easily fixable solution.. It also takes less than 10 minutes of dev job. After all theres plenty of them to do stuff AT submission stage. So as a general rule you should allwatys bake down as much of the animation to simple look ups before sent to frarm in any case. Thsi amkes renders faster and more stable

Its motion capture data, so its only about sticking stuff to the ground.
I admit to writing some very ugly code in my life because i was sick of disscussions
Thanks for all your help.

[joojaa]

QUOTE
How would i use the software renderer through the api?

You dont* just make a dummy light and it allways does this a sa prerender step just tell maya what the name of the damp file should be in the disk based demap section.
QUOTE
Its motion capture data, so its only about sticking stuff to the ground.
I admit to writing some very ugly code in my life because i was sick of disscussions

Then just bake the animation befroe sending to server. MUCH faster,

In any case your really over engineering here.

*command is domain of MEL

[katisss]

using a light thats a good idea, would probably not have got that one.
Just out of curiousity any way to use the software renderer from api?

[joojaa]

QUOTE(katisss @ 06/18/09, 03:44 PM) [snapback]309737[/snapback]
Just out of curiousity any way to use the software renderer from api?

Yes call mel.

Api deals with how things do things in the dg. Mel deals with commanding those things. Theese 2 things dont ooverlap its the job of mel so you should use mel. But fortunately you can call mel form within the api.

Offcourse you can sample things in other ways.

[katisss]

QUOTE:
Offcourse you can sample things in other ways.

I get how to do this with meshes etc. but can i get depth information?

[rob-bateman]

QUOTE(katisss @ 06/18/09, 06:10 AM) [snapback]309741[/snapback]
but can i get depth information?

just render a depth pass.

[katisss]

Id love to, but how do i do it in the node.
I would prefer to not save the file on disk but rather keep it in memory straight.