With the advancement of Nvidia Machine Learning technology, would it be possible to train ipi to recognize where parts of occluded limbs would go? This will come in handy as I won't need as many cameras to get a good solve. Some examples are

1) Using props (wooden rifle gun)
2) multiple actors interacting with each other.
3) Traveling to a location and needing to use less cameras.
4) Bad lightning conditions with harsh shadows.
5) Less than ideal outfits

Blender GURU provides a good examples in this video.

https://youtu.be/FlgLxSLsYWQ?t=938

As a user i also think that it's a must have technology and it will happen since there so much money and utility to it.
It's a huge leap foward and ipi will have to catch up on that one day or another if they want to stay in the competition so i'm not concerned about that.

Ipisoft mentioned they are looking at AI motion tech to integrate into ipisoft, which is great news. Here's an open source one called OpenPose.

https://www.youtube.com/watch?v=vTC0QKR_uM0

I would love to see ipisoft implement this tech and also add support for multiple cameras :)