RC Demo Reconstrution always crashes at "Creating Model" stage at 90% completion

As the title says, its always - I mean always, at 90% completion. Please see below screenshots when using two different image sets.

And here’s the actual error dialog box - unhandled exception error:


Is it possible tell what’s really happening and what RC is doing upon reaching ~90% completion when doing reconstruction? I’m only using Normal mode with “Force single part mode” to True (High mode is taking a veeerry long time). I wanted to have a single mesh only. Below is a screenshot of my Reconstruction Settings:


Its always at 90%. Its not 75% or 98% or anything else. It seems like RC is doing something new and different on this stage that’s probably causing the crash.

Is it a problem with my RAM? This 90% crash thing only happens if I set the Maximal vertices count to 70M and above while anything below like 65M or 50M and reconstruction succeeds.

Here’s my PC specs:
CPU - Ryzen 2700x (not OC)
Board - MSI B450 Tomahawk
RAM - 64GB (16GBx4 3200 G.skill TrydentZ)
GPU - MSI RTX 2070
PSU - Antec Truepower 650W

Please let me know if you need the dump file, will gladly upload it too.

Just to add, I’m using Samsung SSD Pro 256GB for the cache location.

Hi JM,

missed your post, hence the late answer…

I’ve never had that issue but I also never used “Force single part” before. Is that really necessary? You could also let the mesh be calculated at the maximum resolution and then simplify it. The algorythm for that is actually quite good imho.

Oh, and yes, it might well be a memory limitation if it cuts out at a certain size…

Based on my tests earlier, looks like single mesh output is a little faster compare to having multiple parts.

Well, I guess I will just be using multiple parts because if I use multiple parts, mesh reconstruction is fine and I’m getting around 200M tris of mesh.

Yeah, the simplify tool is great and very helpful.

Hmm, interesting.

Can you put a rough number on that? Is it marginal or rather significant?

I always thought that it just does a normal reconstruction and then a simplification in one step as opposed to do it separately, but I’m by no means certain of that…

Its minimal, like 1 or 2 minutes tops on my case, but that was only using 50 photos (image set of a rock at 42mpx) and I’m assuming that if I use a higher number of photos, the speed difference will also increase - well I’m not really sure on this part tbh.