Lets say that i have a object with 300M polygons. My target is 5M polygons
would i get better result by simplifying in smaller steps towards 5M polygons
or
do i get same result if i go from 300M directly to 5M polygons.
Hi janosh,
why not just try it out? 
I am pretty certain the results will not be identical, but very close.
Although reducing fromm 300 to 5 might have severe repercussions on the texture quality…
In RealityCapture in one step.
I tried this and the quality is pretty much the same if you go one step or with smaller steps.
The texture quality in 5M model is totally acceptable. In VR you cannot have much more anyway. with good machine you can run 10M polys in vr but the visual difference between 5M and 10M is something that normal user cannot detect if the model is a room for example. I wish if RC could do better UV’s. They do work perfectly but they are a bit heavy on realtime since they add extra vertices to UV island edges
For objects i do the normal mediumpoly model and new UV’s and bake height + normal + delighted albedo and sometimes I do roughness map by hand. Lowpoly projection in RC doesn’t work well at all but it works really well in Substance designer and in Knald. You can make really nice and super fast UV’s for triangle meshes in 3D coat.
It would be really awesome if I could run 300M model in VR in future we can
Somebody said that if yo simplify in Meshlab you should go in smaller steps to get better end quality. The problem is that Meshlab hangs if you try to import even 40M polygon mesh.