Hi,
Are there any plans to incorporate the AI version of Graxpert into Siril?
1 Like
For now. No.
I want to test if it is relevant. Currently our background extraction works well. Do we want to add another action with another external program.
I’m not sure.
My reason for asking was I was under the impression that the RBF code came from GraXpert team as part of v1.0.2
Yes. They coded directly in Siril.
However for AI this is not as easy. AI is essentially developed in python like GraXpert. But this is not our case in Siril.
1 Like
I’d very much like to see GraXpert integrated into Siril as well, as Starnet++ is.
Pixinsight does this now. Just install GraXpert, point Pixinsight to where it is and off you go. The only thing needed in Siril UI-wise is a dialog with with a drop down for the correction method and a slider for setting the smoothing.
Hello.
I’m not sure this is an argument.
Siril currently uses GraXpert’s gradient extraction algorithm, which works very well. The main difference with the AI version is that it requires samples to be placed. I don’t think there’s enough difference to integrate interaction with third-party software. This is always a source of problems.
And placing samples correctly isn’t as trivial as you might think, especially as a beginner. What is the background and what is faint nubulae? Having AI do the heavy lifting, as an option, would certainly help and most likely surpass human sample placement in many scenarios.
To me it’s a no-brainer having SIRIL send image data to GraXpert and then importing the result back in, but that’s me. When GraXpert improves, so would SIRIL.
it’s not a no-brainer because we have to manage the link for at least 3 different OS and as a GTK+/glib program, this is all but simple. Although we have done that with other programs now and are starting to have some experience with the thing, it will still take time to develop, test and above all, help users who can’t make it work.
Alright, so the no-brainer comment didn’t mean it doesn’t take brains to make it work. Obviously it takes effort. I sincerely apologise if that’s the way it came across.
What I meant is that this is relatively low hanging fruit in order to improve SIRIL and keep it a superb alternative to Pixinsight.
Do what you want with this input. 
Hi!
It seems to me that it is not possible to extract the background with GraXpert AI on each of the frames of a sequence with one command, but only on the stacked image.
Could this subroutine be somehow implemented in Siril?
Thank you!
Hello. No it is not possible because graXpert AI id not integrated in Siril.