IHDM priors

Inverse heat dissipation model (IHDM) priors used for Temporal-Matched Data Consistency (TMDC) on linear inverse problems (deblurring, super-resolution, MRI). One subfolder per training set:

Subfolder	Training data	Resolution	File
`ffhq_256/`	FFHQ	256	`ihdm_ffhq256.pth`
`church_256/`	LSUN Church	256	`ihdm_church256.pth`
`afhq_cat_256/`	AFHQ cat	256	`afhq_cat256_50m.pth`
`imagenet100_128/`	ImageNet-100	128	`imagenet100_128.pth`
`fastmri_knee_320/`	fastMRI knee (magnitude)	320	`fastmri_knee320_100m.pth` (~~100M), `fastmri_knee320_50m.pth` (~~50M)

from huggingface_hub import hf_hub_download
path = hf_hub_download("Lin-Wang-NUS/IHDM", "ffhq_256/ihdm_ffhq256.pth")

The matched ADM (hot diffusion) baselines trained on the same data are in Lin-Wang-NUS/ADM.

Downloads last month: -; Downloads are not tracked for this model. How to track

Inference Providers NEW

This model isn't deployed by any Inference Provider. 🙋 Ask for provider support