Hi Y'all,
I did try to write a spec for smoothing that doesn't require some kind of rounding/snapping process, and I was not able to do it.
Meanwhile, with actual code and actual parts, adding rounding (even poorly written rounding to just 'get started') addresses almost all of the normal errors not induced by T junctions. So I feel like maybe there aren't that many other shoes to drop. :-)
But at the same time, I still feel that fixing T junctions in-app feels weird to me - my gut says I'm writing to code to fix parts that aren't authored up to standard (albeit maybe only a standard in my imagination :-) as opposed to compensating for the limits of floating point.
The minifig faces present a problem - to impose triangulation on the library, we would have to induce into the 'rest of the head' parts enough triangles to ensure uniform connectivity to a variety of actual face pieces, all of which may have different current edge triangulations.
For examlpe, 3626bpa2.dat contains a head suture whose triangles go right up to the 'cut-over' line to the primitive that builds the standard forehead area. The result is a T junction that would require modifying the primitive.
Sergio, your note on caching is well taken - I currently have an algorithm that's tolerable in real-time, but if I have to add T-junction removal that's probably going to go out the window. :-(
(Am I the only one going 'this would be so much easier if the minifig heads were texture-mapped? ;-) Also, it blows me the level of detail people have built colored triangle meshes for so many faces.)
cheers
Ben
I did try to write a spec for smoothing that doesn't require some kind of rounding/snapping process, and I was not able to do it.
Meanwhile, with actual code and actual parts, adding rounding (even poorly written rounding to just 'get started') addresses almost all of the normal errors not induced by T junctions. So I feel like maybe there aren't that many other shoes to drop. :-)
But at the same time, I still feel that fixing T junctions in-app feels weird to me - my gut says I'm writing to code to fix parts that aren't authored up to standard (albeit maybe only a standard in my imagination :-) as opposed to compensating for the limits of floating point.
The minifig faces present a problem - to impose triangulation on the library, we would have to induce into the 'rest of the head' parts enough triangles to ensure uniform connectivity to a variety of actual face pieces, all of which may have different current edge triangulations.
For examlpe, 3626bpa2.dat contains a head suture whose triangles go right up to the 'cut-over' line to the primitive that builds the standard forehead area. The result is a T junction that would require modifying the primitive.
Sergio, your note on caching is well taken - I currently have an algorithm that's tolerable in real-time, but if I have to add T-junction removal that's probably going to go out the window. :-(
(Am I the only one going 'this would be so much easier if the minifig heads were texture-mapped? ;-) Also, it blows me the level of detail people have built colored triangle meshes for so many faces.)
cheers
Ben