Tangent Space From Uv Coord

[pipoun]

Hello!

I looking for a way to find the tangent space from UV coordinates.

In the docs, I've only founded one way:

• MFnMesh:: getPointAtUV ( int polygonId, MPoint &toThisPoint, float2 & uvPoint, MSpace::Space space, const MString *uvSet, float tolerance )
  But it looks like I will have to iterate throught each polygon until I find a MPoint which is inside the polygon. Isn't it a too heavy process?

• After that I will have to compute the barycentric center

• MFnMesh:: getFaceVertexTangents ( int faceIndex, MFloatVectorArray & tangents, MSpace::Space space, const MString *uvSet )
  MFnMesh:: getFaceVertexBinormal ( int faceIndex, int vertexIndex, MVector & binormal, MSpace::Space space, const MString *uvSet)

• And then compute the Tangent/Binormal using the barycentric weights computed previously.

Is there any other way to do that? This solution looks to blow too much processor time!

Thank you for your help!

[rob-bateman]

QUOTE(pipoun @ 08/20/07, 07:56 AM) [snapback]271946[/snapback]
Hello!

I looking for a way to find the tangent space from UV coordinates.

Pardon my stupidity, but isn't that impossible? apply the same texture to 6 faces on a cube and you have 6 possible points relating to that UV and mesh? In this simplistic case, the bi-normals and tangents would all be different for each point? Therefore surely you should be thinking about getting the set of tangents and bi-normals at a given UV?

QUOTE(pipoun @ 08/20/07, 07:56 AM) [snapback]271946[/snapback]
In the docs, I've only founded one way:

• MFnMesh:: getPointAtUV ( int polygonId, MPoint &toThisPoint, float2 & uvPoint, MSpace::Space space, const MString *uvSet, float tolerance )

Could you then use getClosestNormal(), generate two points

P0 = (PointAtUV+normal*scalar)
P1 = (PointAtUV-normal*scalar)

where scalar is some relatively small number?

then use MFnMesh::intersect() to get the faces?

[joojaa]

do you want to do this render time or outside render time? I mean when you lookup values in uv cords at render time the surface allready KNOWS the normal and bi normals! So even if you wanted to do a converter you should use render context to solve the result!

Uvs off course don't work the other way around very well unless you can guarantee nonoverlap,

[pipoun]

Thanks for your help!

QUOTE
Pardon my stupidity, but isn't that impossible? apply the same texture to 6 faces on a cube and you have 6 possible points relating to that UV and mesh?

Rob, you're far from stupid . Actually I hadn't mentionned that there's no overlapping. It's a given constraint.

QUOTE
Could you then use getClosestNormal(), generate two points

P0 = (PointAtUV+normal*scalar)
P1 = (PointAtUV-normal*scalar)

I'm sorry but I don't get something.
Do you want to use getClosestNormal() then intersect() to find the MPoint on the Mesh from PointAtUV? The result won't it be PointAtUV itself?
The problem with getPointAtUV() is that I have to iterate on each Face until I get the good one.
Or I am missing something?

QUOTE
do you want to do this render time or outside render time? I mean when you lookup values in uv cords at render time the surface allready KNOWS the normal and bi normals! So even if you wanted to do a converter you should use render context to solve the result!

I want to do that outside rendertime. But actually I don't care, I'm just looking for a method to get the tangent space for one UV coord. I have a set of UV coords, some thousands, not ordered, and I want to get tangent space for each one.

You seem to say that I can get more information at render time. May you give me some more info about that?

Thank you again!

[rob-bateman]

QUOTE(pipoun @ 08/21/07, 12:24 AM) [snapback]272008[/snapback]
Do you want to use getClosestNormal() then intersect() to find the MPoint on the Mesh from PointAtUV? The result won't it be PointAtUV itself?

It returns the face indices.....

[joojaa]

QUOTE(pipoun @ 08/21/07, 11:24 AM) [snapback]272008[/snapback]
I have a set of UV coords, some thousands, not ordered, and I want to get tangent space for each one.

this may sound stupid but

If your working outside render scope then theres really no use for this. What you do is, you convert the uv to a vertex selection and interpolate on that vertex or face. Polygons are are easy since they are lineary interpolated in value between vertices's.

in a way theres no data outside the vertices outside render context.

PS. it does not matter after fetching whether you interpolate in uv or world space the result will be the same. so you can attcieve same result by searching the values of normals/binormals to the uvs and interpolating on the uv triangle but doing it directly the mesh sace has advantages as you have more functions to help you

[mario-baldi]

A lot of time has passed since this post, but, have you solved this problem ?

I am really going crazy with this, and I don't know how to solve it

[joojaa]

Dont hijack other peoples threads! Start a new one.

because  a) your problem is not necessarily the same you just think it is. b) the original poster may not be around anymore. c) this sites posting system has a tendency to just loose your post, as there's no good way for most users to see other than newest posting date, and 5 latest activities.

a) is more important, so you need to start over and describe your circumstances. c) is something to think about, you were allmost totally lost to everybody. since you were at spot 5 when i answered you (had one other post come most people in the world would have missed your post. Yes its a slight bug)

Again so you want to do it in render context or not. If render context WHAT context mentalray, maya software, hardware rendering, renderman, final render, turtle or what?

[mario-baldi]

Hi Joojaa, sorry I am new here and I thought that clicking on the reply button I would have replied to the entire thread.
About starting a new one, I have a very similar situation to pipoun
I think not to be off-topic, furthermore I hate when, making a search, I found a lot of threads with the same argument.

Anyway what I need to do, is to get a world point position with maya api for custom node, in the fastest way possible.
Actually pipoun had to do some more complex job, but  his first point was to get a point from uv info, which is an heavy process with known maya api methods.

I don't need to know this position for rendering purpose, but to build some vertex geometry starting from uv coordinates.

Thank you for your answer

[joojaa]

About starting a new one, I have a very similar situation to pipoun

Nevertheless. The problem is actually quite different.

but  his first point was to get a point from uv info
I found a lot of threads with the same argument.

that is because this is hard, theres no easy way to do this. Since uv cordinates do not a one to one mapping. Anyway Bateman describes quite exactly how you would do this. He even gracioulsy points out the relevant code. Its just that unless the situation is special its not possible to walk the tree this way.

So what you do is you find out the MFnMesh:: getPointAtUV to get the index, then assuming uvs do NOT overlap or you have some external data or intend to draw many points. You now know what face your targetting. Then you need to find out the barycentric cordinate of the uv space of said tesselated triangle of face this you can use to interpolate the coordinate as outlined.

Since your building geo its probebly easier to just work on edges tough this simplifies things quite a bit.

but to build some vertex geometry starting from uv coordinates.

That makes all the difference. You can now sidestep the entire problem in one batch action.

However! You dont actually need to wrok in uv space at all, you dont actually need to know the mapping form uv 2  to do this as you can use a indexing trick to THINK of the uv space as a 3d space which happens to work for splitting and such very well.

Secondarily you can allready do the same thing with exisiting maya nodes (multithreaded optimized i doubt your api code will beat that). you just need to split,warp the object to its uv coordinates, flat, use normal maya 3d tools and wrap it back, merge. So you can do the same in your node this saves you a lot of time trying to mine data up and down in structure thats ill suited for your probloem.

Or even simpler just make new object in uv space and use the second wrap to make it up in 3d space. If you serach with the right terms then youll see that ive demonstrated this about 50 or so times.

Offcourse its of NO use for getting the Tangent space.tangent-space-from-uv-coord

[joojaa]

Oh PS it doenst happen to be a nurbs surface now does it? If it does then its REAL easy

Anyway im going to close this thread now and force your hand. New thread, you can easily reefer to this thread with a link instead if you think it helps. It keeps tracking issues easier.