[hatts]

Can anyone recommend a decent viewer for very large meshes (15gb+)?

We need something that can view large meshes with very basic shading. Features like clipped views and multiple mesh support would be good too.

So far the Clarisse demo is the only program that allows us to manipulate large models smoothly in real time (which it does eerily easily) but we'd like something less compositing oriented. Polyworks is powerful but just too clunky.

[Mihai]

Meshlab? Haven't tried it at all but it's used for merging scanned meshes with M.A.N.Y triangles.

[polynurb]

maxwell fire?

just thinking,..if you have much ram and find a way to bypass openGL viewport...

[hatts]

Meshlab? Haven't tried it at all but it's used for merging scanned meshes with M.A.N.Y triangles.

We do use Meshlab but the viewport slows down a lot if the model's big enough

maxwell fire?
just thinking,..if you have much ram and find a way to bypass openGL viewport...

Unique idea...unfortunately re-voxelization takes too long in C4D every time we want to hide/show/move parts, etc. Studio would be an option but it can't generally load huge files.

I don't know what secret sauce Clarisse has injected into their code but it's pretty miraculous, just unfortunately not a useful application overall for us.

[polynurb]

clarisse is not based on openGL viewport asfaik.. but not sure never installed it.

in rhino you can disable "redraw" or you could probably set up a display mode that hides everything, because i think gpu ram is the culprit with such large meshes.

i had exterior scenes with thousands of high poly trees that i could not navigate in the rhino viewport anymore but fire handled it ok.

btw. there is a basic limitation with cinema and fire revoxelization that is not present in rhino.. so clipping planes and moving geometry should be fine in rhino.

[Polyxo]

I think none of the previously mentioned apps will be of help, they will not even remotely be able to deal with these massive files.

The only process I know, which output that insane volume of data is CT scanning – yeah and maybe conventionally 3D scanning of an entire
lorry in just one chunk . CT-scanners have their own viewers but they are of course optimized towards looking at small chunks at a time - which is
fine for medical applications. I would be very surprised there was an app at all which lets you see 15 gb of mesh data in the viewport, without culling
95% of the information. Given you're talking about CT data, there's specialized processes to strip out just the surface mesh, this would yield a
tremendously lighter model, which still looks as detailed as the source data.

[hatts]

in rhino you can disable "redraw" or you could probably set up a display mode that hides everything, because i think gpu ram is the culprit with such large meshes.
btw. there is a basic limitation with cinema and fire revoxelization that is not present in rhino.. so clipping planes and moving geometry should be fine in rhino.

Haven't tried the Rhino approach because of it's typically sluggish polygon abilities but with redraw disabled there might be a possibility. The C4D limitation is why I briefly mentioned Studio, but I guess that opens the option of several other CG apps, as long as their respective viewport displays are somehow circumvented.

The only process I know, which output that insane volume of data is CT scanning...I would be very surprised there was an app at all which lets you see 15 gb of mesh data in the viewport, without culling 95% of the information.

Excellent guess Polyxo, most of the data does come from CT/other scans.

We have methods to view the raw scans, but to us the useful visualization is of the converted polygon mesh, not the CT data. Our aesthetic/engineering edits are done in conventional 3D applications, so we need to view our meshes after these edits.

Clarisse indeed displays files over 10GB with ease. This video is titled "CPU powered 3D view" so there's a hint. It then goes on to display realtime feedback of 4.9bil polygons.

[Polyxo]

Clarisse indeed displays files over 10GB with ease. This video is titled "CPU powered 3D view" so there's a hint. It then goes on to display realtime feedback of 4.9bil polygons.

Have you already tried things out with your actual scan-data?
What they show in this promo video are HiRes versions of CG-meshes. As these items are so logically constructed Clarisse can
likely use a lot of optimizations (displaying stuff which is farther away from the camera with lower subd-levels etc). None of the single
elements I see in this clip has a detail level which requires more than some hundred thousand faces and then they get instanced
all over the place, which can also be done quite efficiently. What they show in this clip is certainly very nice but nothing completely
out of this world performance wise.
I could well imagine that the situation was very very different when you gave that program pieces to chew where each chunk has
hundreds of millions of unordered triangulated faces – content the program isn't optimized for.

Edit: I understood that you would want to view raw, triangulated meshes. If your edited versions got also got remeshed /retopologized
to subD-meshes, the situation probably was quite another.

[hatts]

If your edited versions got also got remeshed /retopologized to subD-meshes, the situation probably was quite another.

Of course the meshes come out very nasty when converted from CT, so retopologizing is mandatory. There's no SubD usage but the topology is clean.

[seghier]

the best software to view large meshes is zbrush

[rusteberg]

bob seghier may be right. z - tooth brush can in fact handle large amounts of polys. and it does in fact have some good retopology tools built in considering you bring in a "clean" "mesh"...... but it sounds like you are dealing with a large set of points which are non quad and that might present a problem or five....

it also sounds like you enjoy toying with us based on your profound knowledge

[hatts]

Haha, we do have a weird workflow rusteberg.
zBrush actually has an unofficial file size limit around 1.5gb, at least in our experience, and with OBJs.

But actually zBrush is vital for us, both for sculpting and for retopology.

Anyway the search continues...

[Mihai]

But why do you need to view these meshes at 100% detail of the scan? This is the part of your workflow I don't understand.

If it's ment for retopolizing - then you will lose a lot of tiny detail anyway, you can easily filter the scan to 90-95% of the original and still retain all the major/minor features.
If it's ment for 3D printing - are there really any 3D printing methods that will retain all the detail of such a high detail scan?

So why not first filter the mesh?

Just speculating but the only reason to have such high details is more for scientific analysis, not visualization purposes. Or are these scans of buildings or huge things at 1/10mm detail level?

[hatts]

But why do you need to view these meshes at 100% detail of the scan? This is the part of your workflow I don't understand.

Our goal is to acquire the maximum detail that can be captured via digital fabrication methods, so we're going to use data that's as raw as possible. Believe me, many of my headaches would be cured if filtered/(highly-)decimated versions were good enough.

If it's ment for retopolizing - then you will lose a lot of tiny detail anyway, you can easily filter the scan to 90-95% of the original and still retain all the major/minor features.

Our scans are sometimes objects that are 3cm tall, and scaled up to be 3 meters tall. So a "minor" detail at 3cm becomes a very noticeable detail at 3m.

If it's ment for 3D printing - are there really any 3D printing methods that will retain all the detail of such a high detail scan?

It gets CNC milled, which can achieve absurd detail

[Mihai]

And what's the biggest piece you've CNC milled?

Our scans are sometimes objects that are 3cm tall, and scaled up to be 3 meters tall. So a "minor" detail at 3cm becomes a very noticeable detail at 3m.

But then I would think the 3D scan itself becomes the limitation....meaning, it doesn't matter if your scan is 20GB of data, or 20GB of data cleverly filtered to 3GB, if the original scan resolution doesn't give you anything correct beyond a certain level anyway. It's not micron level we have reached yet with scanning, have we? Plus I would think 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. Just curious but what do you use for scanning?