StarNet++ is a very interesting project by Mikita Misiura.
It's a neural network based software to remove stars from astronomical images, and it does surprisingly well.
This piece of software is so interesting that has been released as an official PixInsight process starting from release 1.8.8-6.
StarNet is very easy and straightforward to use: it has a single parameter, the stride, whose default value is almost always very effective.
The only problem with StarNet is that it can only be used on non-linear images.
In my typical processing workflow I keep my image linear as long as possible and, many times (mainly for Narrow band image processing), I would like to handle starless images still at linear stage.
This is why a few month ago I recorded and published a very simple video-tutorial on YouTube ( https://youtu.be/GboQWoRovJs ).
The technique was very easy, but with some problem on images with a very bright background.
So I worked on PixelMath formulas to get better results.
Finally I decided to turn all this idea into a Script making the whole process very easy to use.
The script name is LinearStarNet and the user interface is very straightforward.
The script source code can be downloaded from the download section of this website o by clicking the following link