News for hackers. Stuff that matters.

A comment on Diffusion Models Beat GANs on Image Synthesis.

Re:Diffusion Models Beat GANs on Image Synthesis by colordrops on Tuesday May 11, 2021 @11:09PM (#27126169)
I don't know enough about this to understand why their model is better, but the imperfections in their images are much more unsettling that other algorithms.
[ Parent | on HN ]
Re:Diffusion Models Beat GANs on Image Synthesis by brink on Tuesday May 11, 2021 @11:23PM (#27126261)
Yeah, some of those images are nightmare material.
[ Parent | on HN ]
Re:Diffusion Models Beat GANs on Image Synthesis by tobr on Wednesday May 12, 2021 @01:59AM (#27127104)
I wonder if it would be possible to train them to specifically avoid generating images that look scary to people. Eg prefer blur or other artifacts over distorted bodies and faces.
[ Parent | on HN ]
Re:Diffusion Models Beat GANs on Image Synthesis by ec109685 on Wednesday May 12, 2021 @04:41AM (#27128124)
The last set of images has some really scary generated humans via the GAN they were comparing against in the male w/ fish examples.
[ Parent | on HN ]