Testing stacking comet images - StarNet2

[Jordi]

Hello Han,

I’ve just tested the new ASTAP comet stacking feature. The idea of using StarNet2 is very interesting; however, my first test didn't work out as expected.

I followed the steps in the documentation, but I encountered an issue. For context, I used the DirectML package to enable GPU acceleration, but it still took a long time to stack 27 images—it seems very computationally intensive.

Here is the procedure I followed:

Selected Ephemeris Stacking and enabled the StarNet2 option.

Loaded the comet light frames in the Lights tab.

Loaded the darks, flats, and flat-darks in their respective tabs.

Set the σ factor to 2.0 in the Stack Method tab.

Clicked Stack.

As a result, I only obtained one output file (the combined image named _NEBULA*) instead of the expected three (the final combined image, the comet/nebula image, and the star image).

The final image appears to have stacked the stars correctly, but the comet nucleus (and likely the coma) is trailed along the ephemeris trajectory.

I've uploaded the files to Google Drive in case you'd like to inspect them. Due to file size limits, I’ve only included the first frame, the last frame, and the resulting stacked image:
https://drive.google.com/drive/folders/1Vqv6sTdLJdKpYjTzoCGDjSXA_OJgJoxp?usp=sharing

Best regards,

Jordi

[han.k]

Hi Jordi,

For some reason your files are not converted to colour prior to split them by Starnet2. So they are still OSC raw grey. I assume the reason is a missing dark and or flat. That will not work properly. StarNet2 requires reasonable quality for detection. I will add a detection to abort early if dark and flats are missing.

The image is from 2014. The latest comet file from 2026 MPC indicates comet C/2014 Q2 (Lovejoy). I assume it will still reasonable work after 12 years but did you press on the "Analyse the selected file in tab lights"?

Can you share a dark for more testing?

cs, Han

[han.k]

I have made some changes in the code to prevent this to happen without warning.

Development version 2026.09.02:
https://www.hnsky.org/astap_setup.exe

The latest 2026 AllCometEls.txt seems to work for your 2014 images. The comet annotation for C/2014 Q2 (Lovejoy) is in the bottom left corner but since the drift is similar it works and comet should be sharp.

I have create a dummy dark and it seems to work as expected. No new dark required unless you can share all images for testing. Try the development version above. If you need a Linux or Mac version tell me.

cs, Han

[Jordi]

Yes, you are right, Han. I ran a quick test without darks and flats using an old data set from Comet Lovejoy. Now, I am using the new version 2026-09-02, added the corresponding flats and darks, and it ran well. Thanks for the support!

Jordi

[han.k]

FYI, the ASTAP version from today will work better using StarNet2. StarNet2 requires an update. ASTAP will give a warning & abort if you try to use an older StarNet2 version.

cs, Han

[Jordi]

Hello Han,
Thanks for the news. I updated both ASTAP to 10-9-2026 and StarNet2 and checked it.
All well so far.

Jordi