OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
AI Models

Denoising Diffusion Restoration Models (DDRM)

DDRM is a Python research implementation that applies pretrained denoising diffusion probabilistic models to general linear inverse problems, including image deblurring, inpainting and super-resolution, without problem-specific supervised training.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Denoising Diffusion Restoration Models (DDRM) — OpenViST project profile in AI Models.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use DDRM when you need a self-hosted Python research workflow for experimenting with pretrained diffusion models on image-restoration problems, rather than a hosted image-editing interface.

Who it is for

  • Researchers studying diffusion-based image restoration and inverse problems.
  • Developers who want to run or adapt the repository’s documented restoration experiments.

Documented uses

  • Generate restored image samples for documented degradation settings such as deblurring, inpainting and super-resolution.
  • Explore pretrained diffusion models as a method for solving linear inverse problems.
  • Reproduce or adapt the repository’s example experiments using its configurations and model files.
Before you choose

Requirements & limits

The repository is public and specifies an MIT license. Its documented experiments run as Python code rather than through a hosted service.

  • DDRM uses pretrained diffusion models and is described as requiring no problem-specific supervised training.
  • The README points to Conda or Mamba environment instructions and lists PyTorch 1.8 and 1.10 as tested versions.
  • The documented examples depend on pretrained checkpoints and datasets arranged in the experiment-folder structure.
  • The sampling command exposes degradation type, timestep count and noise level as parameters.

Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.

Source-based practical guide

Run a documented noisy 4× super-resolution example

Use the README’s ImageNet 256 example to generate restored samples for noisy 4× super-resolution.

Before you start

  • Python environment configured using the repository’s environment.yml with Conda or Mamba.
  • The required pretrained model checkpoint and dataset files arranged under the experiment directory.
  • An ImageNet 256 configuration and an input image folder if supplying images.

Steps

  1. Prepare the environment and place the required models and datasets in the documented experiment-folder structure.

  2. From the repository, run the command below.

    Documented command
    python main.py --ni --config imagenet_256.yml --doc imagenet --timesteps 20 --eta 0.85 --etaB 1 --deg sr4 --sigma_0 0.05

  3. Find generated images under /image_samples/{IMAGE_FOLDER}; the default image folder is images.

Expected outcome

The workflow is configured to sample noisy 4× super-resolution restorations; output files distinguish original, degraded and restored images.

Guide limitations

  • The README’s example depends on the corresponding configuration, checkpoint and dataset files being available in the expected locations.
  • The README identifies PyTorch 1.8 and 1.10 as versions on which the code was tested; it does not establish compatibility with other versions.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
7 November 2026
Original website title
bahjat-kawar/ddrm · Official homepage title · retained in its source language
How we review sourcesSuggest a correction