This repository contains the code of the algorithm proposed in https://arxiv.org/abs/2308.07983 and accepted for oral presentation at ICLR 2024.
This repository can be installed as a python package by running
pip install -I . on the root folder of this git.
You can try this algorithm either with the available scripts in this project or in the benchopts, with comparison to other algorithms at https://github.com/gabrielvc/benchopts_inverse_problem_diffusion_prior/tree/master
The following table was produced by running the scripts/hugging_faces_models.py with the configurations described below.
The configuration for MCG DIFF is defined in scripts/configs/mcg_diff/default.yaml. Running this script took 3 minutes to generate each image, but
this can be made faster by using parallelization.
| Original image | Observation | Sample | Sample | Changes to config.yaml |
|---|---|---|---|---|
 |
 |
 |
 |
seed = 32, task = deblur_2d |
 |
 |
 |
 |
seed = 15, task = inpainting |
 |
 |
 |
 |
seed = 10, task = sr |