Skip to content

SwinIR GUI: Image Restoration Using Swin Transformer

License

Notifications You must be signed in to change notification settings

EnderPicture/SwinIR-GUI

Repository files navigation

SwinIR GUI

Get started

Install dependencies

  • Install Miniconda
  • Install conda environment with command conda env create -f environment.yml
  • Activate conda environment with command conda activate pytorch1
  • Install CUDA Toolkit 11 (windows)

Run

  • Run the gui with command python gui.py
  • GUI
    • select files
      • select multiple files
      • supported format list is not complete, please let me know if I should add any extensions
    • image preview window
      • only shows the first of the selected files
      • click anywhere on the image to show expanded view and wait for ai enhanced results
    • dropdown #1
      • pre trained model selector
      • realSR L x4 is good for super resolution
      • color denoise 15-50 50 being more powerful but removes more detail
    • dropdown #2
      • tile size selector
      • tile size work with window_size and overlap to create final tile size
        • window_size is always 8 except for de-jpeg model, which is 7
        • for example 50(tile size) * 8(window_size) + 32(overlap) = 432px * 432px tiles.
        • 50 - 90 is good for 8gb vram card
    • run button
      • batch run files through selected model, results in /output folder
    • status text
      • shows program status and progress of image processing

What is this

A GUI built on top of JingyunLiang/SwinIR that

  • provides auto downloading of pre-trained models upon first use
  • interactive preview for selected model
  • batch image processor
  • tile size selector to maximize vram usage without going over

Still early in development. Can be unstable and no guarantee is provided.

About

SwinIR GUI: Image Restoration Using Swin Transformer

Resources

License

Stars

Watchers

Forks

Contributors 4

  •  
  •  
  •  
  •