Live3D: Collaborative Neural Rendering Using Anime Character Sheets

Simple but effective.

CINN looks like a batch-wise pooling.

Reviewers today prefer fancy Transformer or NeRF stuff, but the performance may not be that different from those old-school models with some modifications.

I was curious and clicked on the [Dataset] link and got this:

https://github.com/transpchan/Live3D/issues/1

and here's the paper review notes:

https://github.com/transpchan/Live3D/blob/main/eccv.pdf

which complains that it's too similar to

https://anr-avatars.github.io/

and

https://ai.facebook.com/research/publications/continuous-sur...

To me, this looks like great craftsmanship: They have built a working implementation where others failed. And they found a way to cheaply generate lots of training data. But sadly, research optimizes for novelty, not for usability.

I enjoy that the neural net learned how their arms have to clip through their skirts, pretty incredible. Wonder what sampling times on this are.

Is this down?