Basic usage
On first launch specify ESRGAN/BasicSR main directory path in [Settings] tab. You can also set custom paths for folders that contains input, output images, models and others.
You can do this manually or by drag'n'droping files/folders to application form. This also works for .pth models.
Set MaxTileResolution setting to value that corresponds to the amount of VRAM available (default value is for ~2 GB).
Take your original images and put them in esrgan-path/IEU_input. Ignore LR folder.
For other setting please read Settings portion of wiki.
When ready, click [SPLIT] button. Tiles from input images directory will be created in "esrgan-path/LR". Optimal tile resolution is calculated from MaxTileResolution value (biggest possible). On each split all content of LR folder will be deleted.
You always need to run [SPLIT] before ESRGAN, because it will convert images to png, put them in LR folder, separate alpha channels etc.
Check one or more models from the tree view, select preferred output mode (only two last modes are available if more than single model is selected). Click [RUN ESRGAN] button to start ESRGAN script.
When ESRGAN is finished, results will be in "results-path" folder, "results-path/models" or "results-path/images" folders, depending on which output mode you selected earlier.
To merge resulted tiles back click [MERGE] button.
Notice: Don't delete images from input directory before you finished working with them.