Radek,
I have been following some work on building a live stacking function to run on a Raspberry Pi 4 (see als-app.org/index.html).
In the latest communication from the authors, they have developed a software package to run on the RPi4 under Raspbian. Although in an early stage version, could this be added to Astroberry? (see post #8 in this thread www.cloudynights.com/topic/683836-als-astro-live-stacker-v06/)
I would love to see this happen, would you be able to adapt ALS to support the multiple folder format for stacking frames from different filters on a Mono CMOS camera, as the folder structure that is done by EKOS / Kstars?
Forgive me, if you have already done this, but this would be great eye candy for me while I wait for 3-5 minute exposures to complete and also show me that my progress is getting somewhere!
I think there is a difference between Siril which does complex registrations and stacking of lots of images, and live stacking, I believe the ALS tool just does basic image registration / noise reduction and additive stacking, and I believe that the Pi4 can complete an iteration of that well within a normal exposure time.
The Pi has come a long way since its initial incarnation, and most people who have bought one recently have gone for the 4GB quad-core Pi4 version, and quite a few of those have set it up with a USB3 SSD drive, which essentially gives the Pi about the performance of a reasonable budget laptop.
For information, as far as Ekos is concerned, one of the development plans is to interface with external live stackers like the Guide module is doing currently for PHD2 and Linguider (while I agree that only PHD2 is used really).
The objective is to reduce the complexity of setting up the link between the frame producer and the frame processor and consolidate a common API.
Ekos would possibly also include a very simple stacker to begin with (read "just a pixel-to-pixel average over N frames").
One big issue is, as was already mentioned here, the resources available for computation. It would be relevant to create an "Ekos network" of low-cap machines for this purpose.
This is of course still in discussion, as the main problem raised with Ekos is regressions. We need to tackle this first.
I'm following this topic and already added it to "new ideas" file However, it might be the case that computing power of Raspberry Pi and RAM amount could be a limiting factor.
I would be glad to see some testing results provided by users. If they are fine I will definitely add it to the next release.
I tested with StellarMate the use of ASTAP and its Live Stack feature. The good news is that it works quite well. The bad news is that the user interface under RPi is terribly sluggish. Example is live stack of 17 subs: