You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
This repository was archived by the owner on Feb 7, 2025. It is now read-only.
Based on the work in: https://arxiv.org/abs/2207.00050
Basically consists on allowing for SPADE normalisation on the upsampling branch of the diffusion model unet.
Implement SPADE-norm
Allow for SPADE-norm in diffusion models (requires new input arg to the forward method of the Unet, or the coding of a new Unet allowing for this)
Allow for SPADE-norm in VAE (in case of latent diffusion models, it might be necessary to condition as well on the VAE side). Same 2 possibilities as in previous point: modify original VAE to allow for input segmentation conditioning, or add new VAE.