3D is so very hard to do!

I've been trying to learn more about WPF 3D, so I thought I'd start building a C9Guy by hand. It wont be as elaborate as Adams, but its a lot less code.

So far I've worked out the Positions and Triangle Indices for the body and arms, but it doesn't look quite right. The surfaces are composed of triangles, but where the triangles join you get edges and shadows even on what should be a flat surface.



I think its to do with "normals" but I don't understand what I need to do to make it look right.

If you have the September CTP installed you can view the model here or download the xaml file here

I think the framework should at least include basic primitives like Cubes and Spheres, building even the simplest model from MeshGeometry3Ds is a lot of extra complexity.

Doesn't WPF let you work with *.X or *.3ds meshes/models instead?

W3bbo wrote:

Doesn't WPF let you work with *.X or *.3ds meshes/models instead?

If its anything like DX then loading .X files should be a snap because DX supports .x files natively so you don't have to write your own loading code. .3ds though if I remember correctly, you have to write your own loading code.
 However, I would like to know, WHY THE HELL ARE YOU DOING THAT BY HAND?!?! Make a model with a modeler, then load it!

I'm not sure if this version is different, but the previous one only let you load models from xaml. There were various tools for converting other model formats to xaml, but I'm not exactly skilled in any 3D modelling program either.

I really wanted to understand the basic principles of the WPF framework, rather then spend time learning one or other tool.

The Medical demo showed some basic 3D bar charts made out of cuboids, I don't think you should have to use a 3D modelling tool to create such simple graphics.

eddwo wrote:

My favourite 3D modeller is SketchUp, but its a bit expensive, so I've only used the trial version.

I've tried out SketchUp -> .3ds -> Zam3D -> .xaml -> WPF and it works in principle. I'll have another go in a couple of days.

I still think it ought to be possible to create simple models directly in xaml.

But can someone explain how to use Normals to make a several triangles on the same plain appear as a single surface?

Well, I just wrote an brief explaination of surface normals and how to compute them, but either Mozilla or Channel9 has dropped it, and I don't have time to rewrite it now. Maybe I can do it later tonight. I really hate buggy software.

Charles wrote:

tomault wrote: I just wrote an brief explaination of surface normals and how to compute them, but either Mozilla or Channel9 has dropped it, and I don't have time to rewrite it now. Maybe I can do it later tonight. I really hate buggy software.

If the time it took to be brief exceeded session life, then there's not much we can do. What happened, exactly? Did you receive an error from C9 (server)?
C

Well, it was about three or four paragraphs of text, which is brief for me. I drafted the text in the rich edit window, clicked the submit button on my Mozilla browser and it posted the text of Eddwo's that I quoted, but none of the text I wrote. When I hit the back button to go back and try again, all the text I had written was gone from the rich edit window. There was no error message from the server. This is probably a Mozilla bug, not a C9 problem. I'm using an old Mozilla 1.4.1 (geez, I knew it was old, but not that old) browser, which is what is installed on my linux machine at work, and this is the first time I've tried to write an article with it, rather than just reading C9. I've switched to editing stuff in the HTML window, and it appears to work better than editing stuff in the rich text window.

Well thanks for the effort tomault, I'm sorry I missed your explanation.

My favourite 3D modeller is SketchUp, but its a bit expensive, so I've only used the trial version.

I've tried out SketchUp -> .3ds -> Zam3D -> .xaml -> WPF and it works in principle. I'll have another go in a couple of days.

I still think it ought to be possible to create simple models directly in xaml.

But can someone explain how to use Normals to make a several triangles on the same plane appear as a single surface?

Ok, here's a simple explanation of normals and their use in 3D rendering. Note that this is vastly oversimplified.

The normal to the surface of a triangle is a vector that indicates which direction is "up." Normals have the property that they are perpendicular to the surface of the triangle they belong to. Rendering systems use the normal to indicate which side of a triangle is the outside (the "up" side) and to do lighting computations. Because the normals indicate which side of a triangle is the outside, it is important to make certain that all of the normals on an object point outward, or the object won't look right when it is drawn.

There are actually two kinds of normals that you need to be concerned with: "surface" normals and "vertex" normals. Surface normals are exactly what they sound like: normals to the surface of a triangle. You compute a surface normal by computing the cross-product of the vectors along any two sides of the triangle that share the same vertex and normalizing the result to unit length. If terms like "cross-product," "unit vector" or "normalize to unit length" are unfamiliar to you, you can search for them on the web with your favorite search engine. The right hand rule will tell you which direction the normal points.

"Vertex" normals indicate which direction is "up" or "outside" at a vertex of the model. You compute a vertex normal by averaging the surface normals of all the triangles that share that vertex.

Which kind of vertex gets used depends on the shading model you are using. If you are using a flat shading model, then the surface normals are used for the lighting computations and every point on the surface of the same triangle will have the same color. If you are using an "interpolated" or "smooth" shading model, then the vertex normals are used for the lighting computations. Essentially, the vertex normals are used to compute the color at each vertex, and then the colors at the vertecies are interpolated to compute the color at each point on each triangle of the model. This produces a smoother look than the flat shaded model at the cost of more computations.

Hopefully, this is enough for you to get started.