20 ms·
Scene Representations in a Latent Diffusion Model
- yreg 3y agosummary by OP on reddit: > Researchers experimentally discovered that image-generating AI Stable Diffusion v1 uses internal representations of 3D geometry - depth maps and object saliency maps - when generating an image. This ability emerged during the training phase of the AI, and was not programmed by people. https://www.reddit.com/r/StableDiffusion/comments/15wvz2a/researchers_discover_that_stable_diffusion_v1/ https://www.reddit.com/r/StableDiffusion/comments/15wvz2a/re...