[Polyxo]

It's obvious that you can't reveal much but so far I have problems to understand the process.
You have lots of complex scan data, some of this is volumetric data, some is not. In all cases you
managed to extract the relevant surface data and converted it to precise triangulated meshes.

If the final goal is to CNC-mill, these triangulated meshes, maybe moderately filtered generally should
be the best /most precise input for your CAM program. Complete remeshing, especially to quads clearly
is lossy, even if the result looks "cleaner" on screen.

But there's obviously also the wish to composite the mesh items before physically outputting and to
inspect the outcome on the monitor. You now retopologized these meshes, probably using Zremesher
or dynamesh but the outcome still seems to be extremely highpoly, so that it's a challenge to find a
suitable viewer application.

I wonder: If one only needed the viewer for checking out several compositions why couldn't one then use
a typical CG route by remeshing to low poly cages and using displacement (the result wouldn't cause
performance issues in typical CG apps). Obviously this geometry was display only, one then used the
triangulated meshes for milling.

If the goal however was to not only composite the input meshes but to further transform and fuse separate
items - why would one use a subdivision optimized sculpting software as Zbrush then? Zbrush in all instances
requires complete remeshing (dynamesh) in order to be able to fuse stuff together. If max shape integrity
is the goal that's quite contraproductive - why not instead use a sculpting app which allows aforementioned
operations on the raw triangle base, such as 3DCoat (in "Surface-Mode"). That way one should be able to
limit alterations to the original mesh to a minimum.

[hatts]

This conversation veering a bit but it's interesting stuff so I hope you all aren't frustrated by our weird ways.

And what's the biggest piece you've CNC milled?...it matters what material you're CNC milling in also, which would also have a "limitation" to how much original detail it can hold, depending on how porous and brittle it is.

"Biggest" is hard to say, but some pieces weigh many tonnes. Brittleness doesn't affect resolution. There aren't chipping problems if you're sensitive toward speed and tool bit size. We mill steel/bronze/marble/wood/granite.

But then I would think the 3D scan itself becomes the limitation....It's not micron level we have reached yet with scanning, have we?

Scanners definitely operate at micron level. Also remember our pieces are scaled up, so 5 microns becomes a noticeable deviation if the piece is 50x it's original height.

Just curious but what do you use for scanning?

A combination of CT, white light, and laser scans.

Complete remeshing, especially to quads clearly is lossy, even if the result looks "cleaner" on screen.
But there's obviously also the wish to composite the mesh items before physically outputting and to inspect the outcome on the monitor. You now retopologized these meshes, probably using Zremesher

I should clarify: we make one tri mesh from the scans, using specialized software. Any further retopologizing is just used selectively to fix errors or make certain aesthetic edits. We don't globally retopologize for a tidier mesh, even though I'd love to.

All attempts are made to minimize "reinterpretation" of the original form; this is actually Koons' sort of raison d'être. Of course the first meshing is a re-interpretation: this has actually led us to research options for scan-data-to-CNC workflow, which is currently not possible, but give the tech nerds a few years and who knows...

I wonder: If one only needed the viewer for checking out several compositions why couldn't one then use a typical CG route by remeshing to low poly cages and using displacement...Obviously this geometry was display only, one then used the triangulated meshes for milling.

This is a good idea, unfortunately some of the pieces have several layers of this

I can't imagine displacing these types of forms. We use decimated versions for rendering and general aesthetics, but when it comes to inspecting the final form, we need some solution.

If the goal however was to not only composite the input meshes but to further transform and fuse separate items - why would one use a subdivision optimized sculpting software as Zbrush then?...If max shape integrity
is the goal that's quite contraproductive - why not instead use a sculpting app which allows aforementioned
operations on the raw triangle base, such as 3DCoat

zBrush is not our primary tool; it's used in cases where the original data is not as sacred, or where forms are being created from scratch. We do use other solutions that are able to maintain smooth boundaries and respect topology.

Did find one piece of software called Paraview, which is working so-so.

[Polyxo]

Ok with a bit of extra information/clarification this all makes more sense.
I'd say that one should generally be able to displace these tissue layers but it certainly was a lot
less work to work with the original geometry directly.

[Hervé]

https://beta.autodesk.com/callout/?call ... 4E3085C7D0

[hatts]

https://beta.autodesk.com/callout/?call ... 4E3085C7D0

Nice one Herve!

Testing it out now.

[polynurb]

how is the story going?

just wondering..
if you have such a dense data cloud, wouldn't point rendering be sufficient for display?
point to pixel rendering is much faster than mesh based.
like this:

http://www.youtube.com/watch?feature=pl ... rf-HJ4fBls

[Polyxo]

like this:
http://www.youtube.com/watch?feature=pl ... rf-HJ4fBls

Hehe, was about to post the same...
Some people in another forum seem to be chronically unimpressed but if that stuff really works, I'm not.

[hatts]

Ah yes, the infamous Euclideon...

Point rendering isn't a terribly useful technology for us as we need to visualize the polygon edits we've made in conventional CG programs.

Research into this problem has sort of stalled for me. Been a bit distracted by the CryEngine 3 SDK lately. Art gallery interiors in a game engine are deceivingly difficult to produce...

[seghier]

what about topogun
http://www.topogun.com/about/overview.htm