Why is my model coming out worse in RealityScan (desktop) than RealityCapture?

General Feedback & Requests
,
1

So, I recently upgraded to RealityScan from RealityCapture. I think it’s great that it’s a lot better at making captures into one component, but I’ve noticed that the results actually come out…worse?

Here’s a screenshot from RealityScan. About 194 photos and everything was able to be combined into one component:

image
image889×809 77.7 KB

And here’s the one done in RealityCapture, which was broken into multiple components. Even Component 0 with only 60 photos looks better than the RealityScan version, which uses all photos:

image
image950×586 64.4 KB

And yeah, my overlap isn’t great in this example, but I’ve actually had this problems with successful data sets as well. Has anybody else had issues like this with RealityScan?

2

Hello Aaron,
this could be a result of wrong alignment. If all images are aligned it doesn’t mean that the alignment is proper.
I can see you are using images with GPS information, but the errors seem to be inconsistent for your first model. Can you try to set the camera priors as Unknown and set the Force component rematch under ALIGNMENT/Registration/*Settings/Advanced to Yes and align your data again.

Also, the control points can help in this case.