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.