Practical overview
When it fits
Use Mprnet when you have images to process for one of its three supported restoration tasks and can work with its documented Python and PyTorch setup.
Who it is for
- People applying pretrained restoration models to their own images.
- Computer-vision practitioners who want the repository’s training and testing code for deblurring, deraining, or denoising.
Documented uses
- Reduce blur in images using the pretrained deblurring model.
- Remove rain effects using the pretrained deraining model.
- Denoise images with the pretrained denoising model.
- Explore task-specific training and testing code.
Requirements & limits
- The demo’s task, input directory, and result directory are supplied as command-line arguments.
- Separate pretrained model downloads are listed for each of the three restoration tasks.
- The documented environment uses Python 3.7, PyTorch 1.1.0, torchvision 0.3, CUDA 9.0, and cuDNN 7.5; the README names Ubuntu 16.04 as its tested environment.
- The model is described as a multi-stage architecture with encoder-decoder features, a high-resolution branch, and information exchange between stages.
Check current requirements in the official documentation. Source-based guidance does not mean OpenViST has run or independently tested this project.
Run the pretrained image-restoration demo
Process images for deblurring, deraining, or denoising and save the results to a directory.
Before you start
- Python 3.7 environment with PyTorch 1.1.0, torchvision 0.3, CUDA 9.0, and the listed Python packages.
- A downloaded pretrained model for the selected task.
- Images in an input directory and an output directory for results.
Steps
Create and activate the documented Conda environment with Python 3.7.
Install the listed PyTorch, torchvision, CUDA toolkit, and Python dependencies, then install the warmup scheduler as described in the README.
Download the pretrained model for deblurring, deraining, or denoising.
Run the documented demo command with the selected task name and paths for the input and result directories.
Expected outcome
The demo is configured to process the chosen images with the selected pretrained restoration model and save results in the specified directory.
Guide limitations
- The README identifies Ubuntu 16.04, Python 3.7, CUDA 9.0, and cuDNN 7.5 as its tested environment; it does not establish compatibility with other environments.
- The guide follows the repository’s documented commands and is not an execution test.
Sources & review
Source record details
- Source availability
- Active
- Sources checked
- 18 October 2026
- Original website title
- swz30/MPRNet · Official homepage title · retained in its source language