Okay this should be in computers but I thought it's too technical for most of us on here to really understand what it's doing.
Basically generative adversarial networks are a form of neural network (2 in fact) which sort of compete, hence the adversarial bit.
Anyway some clever bods at the University of California have taken as input videos of people who can dance and then map this onto someone else and make a new video of that person replicating the dance moves (the target).
Well I think it is clever and cool.
youtu.be/PCBTZh41Ris
Okay some glitches but target videos are computer generated. And when there's two subjects... the bloke has a reflection in the window for him.
Last edited by: rtj70 on Sat 25 Aug 18 at 23:44
|