Hi! Some time ago i trained a neural network to be capable of subtracting environmental data from scans. I think it does the job well compared to how much i rushed trough the development.
I’m interested in how important does this approach feels to artists.
It works in seconds on consumer GPUs and in minutes on CPUs.
Currently you can’t try it, i kind of lost motivation to keep developing it.
Since MetaShape has a delighter, i thought maybe CapturingReality would be interested in working together and implementing this in RealityScan…
Then i would go on to make a polished version.