It was mesmerising… in a horrific way. Like people shapeshifting into floating masses of limbs.
This is kinda interesting in that you can see how the AI (for lack of a better term) has taken all it’s training videos of gymnastic routines and averaged them out into a sort of meta routine ala cronenberg. Because the motion of the gymnast is so fast, the averaging the AI is doing has a tendency to remove bits on the edges of the central body area.
The body is always present in the AI video because it’s always central in gymnasts. The limbs and head are often missing/malformed because they spend relatively little time in any given space relative to the gymnast as compared to the body. So limbs have a lower average and are trimmed by the AI as erroneous more frequently.
It was probably a bad idea to train the AI on the work of Jake and Dinos Chapman.
This is probably NSFW but is one of their safer works
Maybe so! Thanks for the intro.
I wonder how many times their work has been compared to that of Bosch.
They did a series based on Goya’s Disasters of War, which is already quite difficult viewing.
That makes sense. The computers heard us all say, “Simone Biles is inhuman! What she does is impossible!” and were like, “Oh, OK.”
No, really – it’s a completely accurate representation of fifth-dimensional gymnasts intersecting with our lowly plane of existence.
This functions both as a new genre of horror and an important research benchmark. Nearly 1/3 of research papers at CVPR this year seemed to be about human pose / motion understanding.
This topic was automatically closed after 5 days. New replies are no longer allowed.