Comet Stacking issue with starnet2 on M1 Mac ARM64
Stacking program, astrometric (plate) solver, fits file viewer
Brought to you by:
han59
Menu
▾
▴
Hi Han,
first - great to see you are adding this new feature for comet stacking with starnet2.
I did install the latest CLI Starnet for Mac ARM64 "starnet2_macos-arm64_2.6.0-0231_COREML_arm64_cli.zip"
This will install the files on the iOS filesystem ->
/usr/lib/starnet2/
/usr/local/bin/starnet2
When stacking my comet images "10PTempel" they were stacked an calibrated as cal images. Everything okay until this step.
But when ASTAP is trying to use starnet2 there is an error. See images attached. There is a dialog where i can select the "starnet2" folder or in my iOS context the "StarNet2_weights.mlpackage" i could select. When selelcting there is an error in the console with "Failure".
I gave it another try in a new stacking round, to select the "/usr/local/bin/starnet2" when asked again, but the OS Dialog let me not choose this file - it is greyed out.
What i did see that there was created an "Light_10PTempel_60.0s_Bin1_183MC_L_gain111_20260712-040354_-10.0C_0100_cal_mtf.fits" but nothing more.
Attached i made some screenshots to sort out the issue. Perhaps you can figure out this issue...
Thank you in advance
Rolf
Here the screenshots.
Hi Meister,
Up to now I have tested it only in Windows and Linux. I jst tested it in my old Mac X86_64. But I get this error when executing in the command-line:
/usr/local/bin/starnet2
Last login: Sat Sep 12 14:22:07 on ttys000
macbookpro@MacBook-Pro-van-Apple ~ % /usr/local/bin/starnet2
dyld: Symbol not found: __ZNKSt3__115basic_stringbufIcNS_11char_traitsIcEENS_9allocatorIcEEE3strEv
Referenced from: /usr/local/bin/../lib/starnet2/libonnxruntime.1.23.2.dylib (which was built for Mac OS X 13.4)
Expected in: /usr/lib/libc++.1.dylib
zsh: abort /usr/local/bin/starnet2
This could be caused by my older Mac but your screenshots indicate that at your system there is problem with latest StarNet2 for MacOS. I will report it at the StarNet2 issues. I will post the link here later.
I will also set the standard path to StarNet2 correctly in ASTAP and link the console output of StarNet2 to the ASTAP log in a new version.
cs, Han
I have contacted the author of StarNet2 via personal message in CloudyNights. It could take a few days before I get an answer. You could try to contact him as well via a CloudyNights forum.
My Mac has macOS 11.7.11 so that will not work but your system should work.
Han
Hi Han, thank you for your feedback.
i can see and start "starnet2" in my own terminal. So it does exist und should run normally. Perhaps there is just an issue why i can't select it in the dialog box when asking for the programm folder.
Attached you see my CLI output for my mac.
I have uploaded an new ASTAP version for macOS. Now the selection of the StarNet2 executable will work better ( /usr/local/bin/starnet2 )
Now all StarNet2 log messages are captured in the ASTAP log. See attached. Can you try the popup menu of tab lights on one file as shown in attached screenshot? Send me a copy of the ASTAP log.
cs, Han
Link:
https://sourceforge.net/projects/astap-program/files/macOS%20installer/astap_M1.pkg/download
Hi Han,
this version works now! I could now choose the right file "starnet2" in the /usr/local/bin/ directory. It had a glitch in the beginning, when the dialog opens, to select the origin of the starnet2 file. the process of stacking stopped and was not working anymore - and in the logs was shown, even i selected the right file.
19:37:15
No StarNet2 found at /usr/bin/starnet2
Install the command-line version of StarNet2 first or select correct path.
19:37:38
Executable selected. Try again.
I had to restart the whole process an it continued with the already calibrated files. but ASTAP thought this were the original files, so the result were strange images with no colour. and files with ...._cal_cal_cal .... in the filename.
But - when i restarted ASTAP, cleared the list, put the original comet files once again in, the process started working. Including the starnet progress etc...
It took a lot of time, but in the end the result was great.
Just one question - can you mask, the comet core (perhaps starnet2 treats it like a star) itself, so it will not be removed by starnet? Or ist this already happending?
Attached you find the messages of the log.
Good it is working. The first path usr/bin/starnet2 was wrong and was already stored in your configuration. The path should be usr/local/bin/starnet2 My mistake, a default copy from Linux. For a fresh start by an other users it will not happen.
For my the test images, the comet center is not seen as a star. Can you share one or more images so I can have a look here and ask the StarNet2 author for advice. I assume if the center is very similar to a star it will be difficult to prevent.
You could attach one sample image to your next posting and after downloading it I will remove it.
cs, Han
Last edit: han.k 4 days ago
Hi Han,
I direct mailed you a download link, with multiple exposures of different comets, for testing the core removal with starnet2.
I have not received a message with link yet. I will have look again tomorrow.
Hi Han,
thank you for your great support.
I made a bigger package with multiple comets and different exposure times, so you can test for comet core extraction with starnet2.
===LINK REMOVED===
Does the starnet2 work with Mono R G B images as well? You did fix the correct alignement of mono R G B images for me in earlier versions, where starnet was not yet introduced.
Please delete the download link after receiving the images.
Last edit: han.k 3 days ago
Hi Meister,
I have downloaded the samples and removed the link.
Using your images, I can see that for small comet heads it is partly interpreted as a star. If the ephemeris is up to date, I can identify the comet head in the star-only image. So in principle, I could remove the comet head, so a section like 30x30 pixels from the star-only image and add the flux back to the comet-only image using my program. This will only work if the ephemeris so the position is known accurately. If the image is a few years old there will be drift resulting on a offset in the position. To fix that you have to use an older ephemeris data . That is also no problem. I have several older ephemeris file and can even recalculate backwards using the planet positions and forces.
I will try to implement the flux correction first in my program.
I will ask if the StarNet2 author can do something similar if I pass on the x,y position to StarNet2 to "protect" the comet.
Mit freundlichem Gruss.
Last edit: han.k 3 days ago
For testing can you share the full series of one comet? E.g. 10P/Tempel.
Here is my email address:
http://www.hnsky.org/hanemail.gif
Last edit: han.k 3 days ago
ASTAP converts the raw's to colour. This to get rid of the checker pattern. Then they are fed to StarNet2. No colour balance or smoothing is applied. Only at the end of the stacking.
In principle the comet flux in the star-only image should be removed by sigma-clip stacking since it is drifting. Does this work for you and does the comet head still look good after stacking and missing flux?