Yeah, haven't tested it but I assume this would work for realistic cartoon/cgi characters?
Maybe the researchers didn't want to freak out anyone who does 2D animation? "Hey everyone, look at this toy we made that would never be used to animate anything but toy doodles. No one's 2D animation jobs are being replaced here - this project would never do that!"
This looks like an algorithm like how hash map is an algorithm.
It’s also being presented at SIGGRAPH which further makes me think it’s just an algorithm to map motion capture data to a 2D image. We can already map motion capture data to a 3D model.
Yeah I didn't read through it earlier. Seems like there are 3 parts from my understanding. You need some motion capture service like Rokoko/DeepMotion to create BVH files describing the movement from a video. The algorithm this guy created animates/renders by taking the BVH and an annotation file from the drawing. But the project did "train a drawn humanoid figure detector and pose estimator and provided scripts to automatically generate annotation files from the model predictions" so that part does have machine learning.
I would love to use this for my old animation website https://www.superanimo.com/ But due to the AI boom I have tons of work to do on my other projects, so I can't sink time into modernizing my crappy animation site.
Look at what AI generated images were like a year or two ago. They were mostly abstract, dream like images that were cool but not usable. They have improved 1000x in a year. Bottom tier animation from cartoon images now, could be seriously impressive in a few short years.
