RigNet: Neural Rigging for Articulated Characters
Sep 07, 2020 by CGPress Staff
6
|
RigNet: Neural Rigging for Articulated Characters is a new research project that proposes a method to predict and generate ready-to-use rigs based on a 3D model. It can even estimate skin weights from the skeleton. Watch a video of the technology on YouTube and read the research in full on the project’s website.
rigging, uv unwrapping, retopo – all of that is soon going to be automated with AI… some may cry but I guess most will shout with joy !
As long as it’s good 😉
This is a funny statement, they’ve been saying this for many decades especially for UV unwrapping! All the algorithms in the world and you still can’t unwrap anything other than some environment background assets perhaps in an automated unclean fashion. Try giving that to your texture artist.
As for rigging, it is even worse! how can anyone automate rigging? At best it will be able to position a pre-made rig in an object, maybe. But at the end of the day no rigger will cry over anything, because you will need proper rigging for everything, something AI will never be able to achieve because rigging is an artform by itself and AI can’t do art. Unless you want to use these algorithms on cheap quickly made models without much thought then yes it may work, they will be the “poser” of the rigging tools, one click rigging solution.
Retopo maybe the only thing that could work out but it will still need heavy artist input to guide it.
i believe you are wrong… look at what pro Go players said before the Alpha Go project… times are changing and sooner or later they will change…
You can’t compare a board game AI to this. AI like this is overrated, there was AI since back in the 80’s playing chess this didn’t mean that in 20 years we would have HAL in 2001.
Don’t get me wrong I don’t mind at all a tool unwrapping my UV’s for me. But its just talk at this stage just like VR was the talk a few years ago and was going to replace keyboards and mice!
These tools are a set of “do’s and don’t” they don’t “think” the larger the neural network of “do’s and dont’s” the better their accuracy but this doesn’t mean they have any form of problem solving skills which is why they will never replace a human user only assist.
I don’t think i’m wrong, I am simply managing my expectations well.
mixamo: the resurrection.