Hello,
I have about 100 photos of an object and when I perform a draft align, the cloud points demonstrate the object, as they usually do. When I do a full align, they skew and become something unusable.
I’m intermediate with photo scanning in general; with the most experience in a competitors program. I’ve been outputting many models in my first few weeks with Capturing Reality, and love the program. First time I’ve run into this issue.
My only comment is there are some reflective areas I can’t negate, but I’m confused why a draft align would seem to work, but a full align wouldn’t.
Thanks for any thoughts!
Update! I found a post with these settings:
https://www.capturingreality.com/forum/download/file.php?id=978
The object does align better now.
Any tips on how to maximize the power of 32gig ram, 16 core ryzen, 11gig gpu? How far can I stress the settings?
Any guesses why that helps - what’s happening different with these settings? Picking up clues to understand how things work in RC.
hello, are you using the default settings? if not, please try to reset them by holding the shift key when running RealityCapture, and try doing the alignment again
The other way to reset (not just while opening RC) - see the screenshot link above - at bottom left, in current versions, there’s a reset button (also buttons to save a settings configuration, and another button to recall a saved configuration)
Hey Andrew,
I think you need to try that out and tell us! 
Tom, we would need to know which settings Andrew used before. My guess is that it’s the features of 80k and preselector of 40k (which means RC uses 40k tie points instead of the standard 10k).
Hello all!
I was using default settings the first time. The model was an airplane, about two feet long. By playing with the image I posted in my second post on this thread, I was able to generate the model. However, in the long run, I hadn’t done enough to differentiate the wings from above and below, so I’m going to reshoot.
My takeaway was that the default settings may need to be expanded to include more point generation; it’s likely (?) that the default settings aren’t too tasking for basic setups, so users can get started easier. I hadn’t had problems before.
This project was an intentionally complicated one, to test working with lighting on objects that aren’t ideal.
I’ll be reshooting on a greenscreen today, and trying to put differentiating marks on the airplane wings to help the software recognize the wings exist. Otherwise the similarity is likely to cause issues again.
Can you clarify what way you suspect that your airplane isn’t ‘a basic setup’ and so may need ‘more point generation’? Wishgranter has a standard suggestion of a revised set of settings, which may be what your screenshot is - but old hands here also say that ‘more point generation’ will probably just mean ‘more inaccurate points’.
As I said above, “Picking up clues to understand how things work in RC”.