> Both would be killer features. Don't dismiss them because the examples are tri based. :)

Don't get me wrong, I don't doubt that they would be useful. But reading a research paper and making a fully functioning production ready implementation are two totally different things...

It's just far beyond the scope of what I will be able to accomplish anytime soon.

Since the technology is not specific to NURBS and works on triangle meshes, why wouldn't you want to have this functionality available inside a polygon modeler where you might be importing a bunch of other triangle-based models already? It seems like it would be a lot more useful and more general purpose if it was implemented over there rather than only as a specialized NURBS export option.

- Michael

I`ll check that with the silo guys :-D

Thnx for the tip Michael,

Thorsten

I still have a bunch of things to fix, but ngons with aligned vertices are starting to work, here are a couple of tests (Steph's cannon and church). There is still some work to do, for instance there are some messed up parts on the interior walls of the church, but this should give a good idea of how it's going to work.

Now ngons that are adjacent to one another will share the same points along trim curve boundaries.

So it would be good to know if this is better than the previous versions.


These have no texture coordinates, I temporarily removed those to reduce file size.

I also removed the text on the floor of the church - several of the letters had some small extra loops in them which interferes with the meshing. If Steph has the original curves for these it would be good to know if the loops were in the original curves, or if they were created due to some bug in a trim or boolean operation.







Edit: removed church one, there seems to be some problems in the original model.

- Michael

Transmited to Steph :)

From Steph
The letters of the floor Church were simple Text curves (format 3dm) from Rhino 3D imported in Moi
Extrude + Chamfer ---> seems that gives problem :)

For the Gun : points are not yet "auto welded", good matching surfaces, all normals are in the same way :)

For the church : here an Avi video with some bugs:) http://stephlx.free.fr/MoI/bugObj.avi 13 meg :)

For the Old Door : after export Obj in C4D no possibility to "auto weld" points of the fence, too numerous points, no enough Ram !!! (2 Gigas:)

Various complaints :)
Infinite grid on the "2D views"
Scale Tool ( I suppose 1/100, 1/1000...or unity cm, m, inch ...) ?
Text cotation (?)

>> So it would be good to know if this is better than the previous versions.

Some good news and some less good ones... ;-)

The cannon :
- No more missing geometry in Modo ! Cool
- Normals of some faces still pointing in the wrong direction (Modo) I had to flip them manually (or use a double sided material).
- Not new, but once the vertices are merged some visual artefacts appear (due to smoothing). Most of them disappear when I change the smoothing angle of the mesh but I can't get rid of some of them (Modo, Hexagon, Vue)

The church :
- Some missing geometry in Modo (mostly pieces of the ground)
- For some parts, it seems that two polygons (or two objects) are created for the same faces, pointing in opposite direction. It's also the case in the Demo version of Cinema4D 9 but it's less noticeable there as it doesn't generate visual "problems" on screen.
- Face with inverted normals as for the cannon
- I could not merge the vertices of the model in Modo and Hexagon. Both programs have frozen and I had to kill them.
- Same smoothing artefacts than for the cannon

Just let me know if this is clear enough. If needed I can send some screenshots tomorrow (or maybe record some Wink presentations to explain the problems I encounter)

Cya

Thanks Schbeurd for testing in Modo.

The church is fairly messed up from my own bugs right now so I want to focus more on the cannon at the moment.

Tomorrow can you please post a screenshot showing some of the faces that are in the wrong direction? I'd like to closely inspect those areas to see if I can figure out what might be happening there.

- Michael

> For the church : here an Avi video with some bugs:) http://stephlx.free.fr/MoI/bugObj.avi 13 meg :)

It seems that the duplicate surfaces are in the original file.

There are also some other loops in trim curves that are causing the meshing problems. These don't seem to be the result of a chamfer, it looks like a trim or boolean between an extrusion and the arch. If Steph has any of these original curves or objects from before the trim or boolean, those could help to track down what is causing those loops. It is hard to figure out what is causing them just by looking at the loop afterwards. If I can get the objects involved _before_ the operation that causes it, it is better.





- Michael

From Steph
Arch are just curves extrude + sculpt with bolleans function
Only "text" were import from rhino 3D
All the rest is 100% Moi

> Arch are just curves extrude + sculpt with bolleans function
> Only "text" were import from rhino 3D

Does he have the original curves before they were extruded?

If I can get the original curves that were used to create the extrusions, then I might be able to repeat the boolean and see what is happening that causes the loops. It is hard to try and figure out what caused the loops by just looking at the final result.

- Michael

> Tomorrow can you please post a screenshot showing some of the faces that are in the wrong direction? I'd like to closely inspect those areas to see if I can figure out what might be happening there.

Here we go. I've added comments directly in the images (I'm a bit lazy) ;-)

(Edit : forgot to mention that between image 7 and 8 polys were "manually" flipped. Can be quite tedious in Modo with an object made out of multiple parts which is why I prefer to merge the vertices)

Cya

Schbeurd

Uhm, I've been away for a while, and am not sure if this has been brought up before. One sugestion for making the polygon export tool more flexible would be adding sliders for minimum and maximum egde lengths. this would make it easier to avoid the worst of "sliver" polygons.

There's something wrong with the front wheel and front fender, so it might be a good test.

This was the first thing I ever made in 3d (rhino).

> chaos-bike.zip

Hi Bill, thanks for the test model. Here is what it looks like with the current ngon mesher:



The .obj model is a bit large so I'm not posting that right now. I'm hoping to put together a version with this new mesher pretty soon.

- Michael

This is a really cool project. I thought I would bring a quote in from one of the Lux developers so you could read it. "It looks like an interesting project. The developer still seems to be tweaking the exporter, so there doesn't seem to be a link to the latst test cannon .obj. One thing I would suggest to the programmer is to use the .obj 'o' directive to define different object layers. Many obj exporters use breaks in the vertex lines to define layers, but importers that rely on this will not play well with Sketchup export, as you may remember." -Arnie Cachelin

>> ...I thought I would bring a quote in from one of the Lux developers so you could read it...

Is this a thread on Luxology forum ? If yes, can you provide a link ?
Thanks

> Is this a thread on Luxology forum ? If yes, can you provide a link ?

http://forums.luxology.com/discussion/topic.aspx?id=12649

> "It looks like an interesting project. The developer still seems to be tweaking
> the exporter, so there doesn't seem to be a link to the latst test cannon .obj.
> One thing I would suggest to the programmer is to use the .obj 'o' directive
> to define different object layers.

Right now I'm using the 'g' directive to differentiate between different objects.

I'm kind of uncertain whether it is better to use 'g' or 'o' for this... I may need to look at a couple of simple .obj files that contain multiple objects (like 2 cubes) from a few programs to see what is the most common method. But I guess I could put in both 'g' and 'o', that might be the safest...

The exporter is still in the process of being tweaked. I should be ready to release an early version within a couple of days.

Previous versions have had ngons that have some of their points bunched pretty close together, or even duplicated. It seems like this is an area that Modo is particularly sensitive about. Silo and C4D seem to do a better job of either handling or ignoring these. I am currently in the process of improving this though.

But there do seem to be some other problems in Modo as well - it seems as if problems with a particular polygon can then trigger problems with other more simple ones after it. There are a couple of messages way back in this thread where I examined a simple (I think 5-point) polygon that was messed up (along with other more complex ones) when trying to bring in the full mesh, but it was handled ok when it was brought in by itself.

The last test file that I posted had some progress on removing duplicate points, and modo seemed to handle this version a lot better, so I hope that with my final remaining tweaks that also improve point spacing, it should help out a lot too.

- Michael