SinFusion: Training Diffusion Models on a Single Image or Video
abs: https://arxiv.org/abs/2211.11743
project page: https://yanivnik.github.io/sinfusion/
@akhaliq Nice! Irani lab, perfectly in style, does with a single image what others do with many. I am wondering though, how a single large image is different from many small images? Isn't it the number of diverse samples of 'semantic' instances that matter for this?
