I have seen AI form finding using text based prompts, which yield questionable results at this point.
What I am more interested in, is training 3D scan mesh adjustment and feature detection based on before-and-after dataset. Does anybody have experience with this kind of thing? Any resources you can point to?
What kind of objects?
Ae you also doing the post processing of the scans or is that done by someone else?
In Daniel Shiffman’s new “Nature of Code”, in chapters 9 to 11 (cf. link) they go through the basics of evolutionary computing, neural networks, and “neuroevolution”. It’s a great read, if your relatively new to all of this and very approachable.
As I understand it, you’d have to feed it lots and lots and lots of cleansed data, which seems unpracticable for a single person to comb through. In statistical machine learning, garbage in means garbage out.
My current application is a cast for arms and legs. The casts are generated based on manually set points on scans such as the ulna, palm, thumb, ankle, etc.
I’d like to automatically identify those points. I also foresee other ways I could use the ML but this would be the simplest one. The product is being used already so we have hundreds of scans.
