OpenViSTAI · Agents · Blockchain · Crypto
Submit a project
Menu
Image & Video AI

Srn Deblur

Srn Deblur is an open-source Python implementation of a scale-recurrent deep-learning model for removing blur from images. Its command-line program applies pretrained model weights to images in a folder and can also train models from a dataset.

Explore the practical guide ↓

Independent OpenViST project profile, based on linked sources.

Srn Deblur — OpenViST project profile in Image & Video AI.
OpenViST project illustration. See the linked sources for product details.
Understand the project

Practical overview

When it fits

Use it when you can manage its documented legacy Python and TensorFlow environment and need local, folder-based image deblurring or a research training workflow.

Who it is for

  • Developers who want to run image deblurring locally from a command line.
  • Computer-vision researchers comparing deblurring model settings or training on image datasets.
  • Image-processing pipeline builders seeking a folder-based restoration stage.

Documented uses

  • Batch-process blurry images from an input folder and save restored outputs.
  • Select among grayscale, LSTM and RGB model settings for restoration or comparison.
  • Train the model using a dataset and command-line hyperparameters.
  • Evaluate generated color results using PSNR and SSIM measures.
Before you choose

Requirements & limits

  • The documented stack is Python 2.7 and TensorFlow 1.4, alongside SciPy, scikit-image and NumPy.
  • The README supports NVIDIA GPU or CPU execution and warns that CPU testing is very slow.
  • Configured image height and width must be multiples of 16; memory limits can trigger downsampling followed by upsampling.
  • The gray and LSTM models are suggested for restoration and performance comparison; the RGB model favors color consistency over sharpness.
  • The repository lists an MIT license.

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 pretrained deblurring on an image folder

Process images in a folder with a pretrained SRN-Deblur model and save results to an output folder.

Before you start

  • Python 2.7, SciPy, scikit-image, NumPy and TensorFlow 1.4.
  • An NVIDIA GPU or CPU; the README says CPU testing is very slow.
  • A pretrained model downloaded using download_model.sh inside checkpoints/.

Steps

  1. Clone the repository and change into its directory.

  2. Download pretrained weights using checkpoints/download_model.sh.

  3. Run the command below with the input and output folder paths.

    Documented command
    python run_model.py --input_path= --output_path=

  4. Set the configured height and width to multiples of 16; use --gpu=-1 for CPU or specify a GPU with --gpu.

Expected outcome

The program processes the input folder and writes results to the specified output folder.

Guide limitations

  • CPU processing is described as very slow.
  • If the image dimensions do not fit the available GPU memory, images may be downsampled for processing and results upsampled to original size.
Documentation 1 ↗
Source-based practical guide

Prepare and start model training

Configure the documented training command for a dataset placed in the repository's training folder.

Before you start

  • The repository's Python 2.7, SciPy, scikit-image, NumPy and TensorFlow 1.4 prerequisites.
  • A training dataset placed in training_set/; the README says its provided datalist_gopro.txt can be used.

Steps

  1. Place the dataset in training_set/.

  2. Use the supplied datalist or configure the training data list.

  3. Start training with the documented command pattern, for example the command below.

    Documented command
    python run_model.py --phase=train --batch=16 --lr=1e-4 --epoch=4000

Expected outcome

The program is invoked in training phase with the selected batch size, learning rate and epoch count.

Guide limitations

  • The README documents a training command and data placement but does not specify a completed training outcome.
Documentation 1 ↗
Evidence and transparency

Sources & review

Official repository ↗
Source record details
Source availability
Active
Sources checked
17 October 2026
Original website title
jiangsutx/SRN-Deblur · Official homepage title · retained in its source language
How we review sourcesSuggest a correction