More intuitive inpainting

[UserUnknownFactor]

Is your feature request related to a problem? Please describe. Currently inpainting is very difficult to use as is so please make it more availible.

Describe the solution you'd like User actions:

• Use any selection tool to select an area of the image to inpaint
• Call the inpainting plugin, which warns the user if there is no selection availible or layer is empty / non rasterized.

The following actions are done by the plugin:

• Fill the selected area with a solid color without antialiasing on a new (mask) layer;
• (if needed) Convert the mask layer to a format the inference module understands;
• Save image layer and mask layer as cache0.png / cache1.png;
• Send them to the inference module which saves its output as result.png or produces an intelligible error text (The inference module should fill only pixels corresponding to non-transparent mask-layer pixels leaving the rest transparent.);
• Replace the mask layer with result.png so the it will have only inpainted region;
• Delete cache and inference result files.

Additional context Currently if inpainting fails it does so silently without any intelligible error messages whatsoever, please fix it too.

[kritiksoman]

We’ll update the inpainting model in gimp 3 ml branch

[kritiksoman]

Please how line 126 to 160 in

inpainting.py for UI change. Or directly ping on slack https://join.slack.com/t/gimp-mlworkspace/shared_invite/zt-rbaxvztx-GRvj941idw3sQ0trS686YA

[UserUnknownFactor]

Mm.. it's not exactly what I asked. Error checking is also part of it but... Can you at least help me with GIMP3 code for what I asked as the main points? Aka selection, layer generation and filling before sending to the inference module?

[kritiksoman]

In GIMP, see Help->Procedure Browser for list of functions available and it's usage.

[UserUnknownFactor]

It's not that I didn't know about help... I tried it like below and got a lot of type errors and I can't figure out what each type should be by myself:

    pdb = Gimp.get_pdb()

    interlace, compression = 0, 2
    
    drawable = pdb.run_procedure('gimp-image-get-active-drawable', [image])
    arr_drawable = Gimp.ObjectArray.new(Gimp.Drawable, [drawable.index(1)], 1)
    cut_result = pdb.run_procedure('gimp-edit-cut', [1, arr_drawable])
    mask_image = pdb.run_procedure('gimp-edit-paste-as-new-image', [])
    mask_layer_group = pdb.run_procedure('gimp-layer-group-new', [mask_image.index(1)])
    
    pdb.run_procedure('file-png-save', [
        GObject.Value(Gimp.RunMode, Gimp.RunMode.NONINTERACTIVE),
        GObject.Value(Gimp.Image, mask_layer_group),
        GObject.Value(GObject.TYPE_INT, 1),
        GObject.Value(Gimp.ObjectArray, arr_drawable),
# ...
    pdb.run_procedure('file-png-save', [
        GObject.Value(Gimp.RunMode, Gimp.RunMode.NONINTERACTIVE),
        GObject.Value(Gimp.Image, image),
        GObject.Value(GObject.TYPE_INT, 1),
        GObject.Value(Gimp.ObjectArray, arr_drawable),

[kritiksoman]

sorry but there is not much documentation available, the best way is to see other plugin files and see usage.

[kritiksoman]

this is the only documentation available for the time being https://github.com/GNOME/gimp/tree/master/devel-docs/GIMP3-plug-in-porting-guide