RFC: new event after preset or style has been applied

[deekayhd]

I am thinking about a way to trigger some actions after a style has been applied., Typical examples would be:

• rotate and perspective: analyze line structure and correct
• AgX: auto tune levels
• Filmic: Auto tune levels
• ...

My current idea is to execute a Lua script after a style (or maybe preset) has been applied. However, to my knowledge this would need a new event. There is already an event DT_SIGNAL_PRESET_APPLIED, but it is not exposed to Lua scripts. I could not find any similar event for styles.

@wpferguson What is your opinion?

[wpferguson]

What is your opinion?

Think bigger. 😁

You're thinking is

• style -> script
• preset -> script

but the better way is

script shortcut which "evaluates" the image then

• applies a style
• enables/disables modules based on "evaluation"
• adjusts modules using presets
• adjusts individual module controls as necessary

---

Here's what I do...

I shoot sports and portraits so I have a style for each. I also have a general purpose style. The sports style has multiple module instances configured with presets, that are turned off by default.

I cull selecting the images I want. As I pick an image I also assign processing tags such as backlit, over exposed, under exposed, high contrast, low contrast, etc. If the sport was outdoors I assign weather conditions (the script defaults to clear, so I just have to assign if different).

My folder path names will have what sport, the visiting team, and the home team, or portrait, or none of the previous which gets categorized as general.

When I load an image in darkroom the following happens

• the path gets evaluated and the appropriate style applied
• the processing tags and weather tags are used to lookup processing presets in a table and the modules are enabled and the presets are applied.
• If it's an outdoor event I calculate the sunrise and sunset times so that I can then figure out blue hour, golden hour, etc and adjust the modules accordingly.
• If it's an indoor event I apply the color calibration preset corresponding with the venue
• I check the lens camera combination and apply the necessary lens correction.
• I read the pitch and roll data and correct the image tilt.

Then I do adjustments to taste crop/expsosure/etc.

Then I have a shortcut that applies all the "heavy hitters"

• denoise profiled based on camera and ISO
• DorS local contrast, AA sharpening, and denoise depending on ISO
• I increase DorS AA sharpening depending on camera and lens
• A final contrast adjustment based on processing tags/environmental conditions

As a last step AgX is automatically focused so I can make any final adjustments.

Then I move to the next image and the thumbnail of the just processed image is updated in background so that when I return to lighttable I don't have to wait for all the thumbnails to update.

---

Before my camera had pitch and roll data, I automated the rotate and perspective correction using the verticals. It mostly worked, though a couple of times I had to reset and do it by hand.

[deekayhd]

Wow, that is a very clever and elaborate workflow. Now, all you need is an AI model, that automatically assigns the processing tags. ;-) I may steal some of your ideas for my workflow (not the part with the path names, as my shootings are too heterogeneous).

Then I have a shortcut that applies all the "heavy hitters"

Why are these not applied at the beginning or part of the style you apply in the first step?

I am not sure, if this is a standard kind of workflow for many people. My idea was triggered by a post on pixls.us (where else?). My assumption is that most people look at the image and apply some style to it, but what is missing is the click on some buttons for certain modules (I know about the Initial workflow module).

Now, when I think about it, I notice that my idea would probably not work in the lighttable view, anyway. We would need some direct integration into the pixelpipe, not only a replay of shortcuts, that may not be valid in the active view. And switching to darkroom does also not work reliably for me: I often get the message "data pending - please repeat".

[wpferguson]

Why are these not applied at the beginning or part of the style you apply in the first step?

Because they all are applied near the end of the pipe and would get rerun for every change I make early in the pipe. Since they uses lots of resources, they would just slow things down. My early adjustments are crop, exposure, tone equalizer so I don't need to recompute for every change. Once I have the early settings where I want, I turn on the rest and then make final adjustments. Most often I don't need to or maybe I just add a little contrast.

probably not work in the lighttable view

About the "processing" you can do is apply a style or an xmp file.

And switching to darkroom does also not work reliably for me: I often get the message "data pending - please repeat".

There's a trick to that 😄. I really need to write the Darktable Lua Programmers Guide. I'll do another comment with some code attached.

[wpferguson]

action.zip

unzip this and drop it into your lua/lib/dtutils/

In a script do local da = require "lib/dtutils.action"

In your script create a function to do whatever with the loaded image. After that add da.register_events(<name of your script>, <name of function to process image>)

When you open an image in darkroom, your processing function will automatically run after the image has finished loading. You can load an image from a script using dt.gui.views.darkroom.display_image(image)

The library also has short names for all the modules, not including current master, "short" functions for doing actions and da.wait_until_pixelpipe_complete(), which monitors the pixelpipe and let's you know when an action finishes.

[deekayhd]

Thank you so much for this information. Very helpful!
Somehow I always missed the pixelpipe-processing-complete and darkroom-image-loaded events in the documentation. And the helper methods in action.lua are very useful. Will test it later.

[deekayhd]

@wpferguson I think there are two typos in function dtutils_action.destroy_events(script_name):

1. pixelpipe**_processing-complete instead of pixelpipe-**processing-complete")
2. dt.destroy_event(script_name, "pixelpipe-running") <- event does not exist.

I am testing with the current github master branch of darktable.

[wpferguson]

pixelpipe-running was an event that I added for testing then got rid of. I think I commented out the place where I installed it but forgot to cleanup the delete. The code is definitely a work in progress although I do use it everyday. I also haven't updated it in quite awhile. so there may also be things out of date.

If you want to clean it up, go ahead. It's probably good that a different set of eyes looks at it 😄