Connect with the community further in our official Discord server.

Hi everyone!

Tried my luck with the Rosette nebula with my stock Sony A6000. It's about 1,5 hours of data. The hydrogen red signal gets slashed by 75% by the sensor unfortunately, but I still wanted to give it a shot.

Stacked in Siril, then graxpert, then photoshop.

I'll be revisiting this one for sure. Feedback appreciated!

Taken with a Canon EOS 550D with an 80mm lens on an equatorial tracking mount, under Bortle 4 skies. 30 thirty second lights, 50 darks, 50 flats.

Stacked in Siril following the manual pre-processing tutorial. Stretched also in Siril using Asinh Transformation. I set the stretch factor to 30, and the black point to 0.005, then exported the image as a jpeg.

I think she looks a bit better than my attempt at Orion. You can sort of make out the Ring nebula.

Also, I'm pretty sure that's a comet over to the right of Sulafat. Tiny green cloud of consistent size, drifted a few arcminutes over the ~20 minutes I was capturing these images

A really interesting string of galaxies that are part of the Virgo cluster. The interacting pair in center frame are nicknamed "The Eyes".

Dwarf Mini
60s exposures at gain 60
About 5.8 hours worth of sub-frames over two nights
Stacked on the device, post-processing done with Seti-Astro Suite Pro

I've been chasing this one for a couple weeks now. It always seems to be behind trees or clouds whenever I try to capture it, but I finally managed to get a some solid time. I'll get more when I get a chance to try to pull out more deail.
Dwarf Mini, 108 60s exposures, Gain 60, duo-band filter
Post-processed with seti-astro-suite pro

Used the JWST's NIRCAM data (calibration level 3) from filters f444w-f470n, f335m, f227w, f200w, and f187n.

I tried to use the visible range filter too (f090w) but it was too artifacted even with extensive cleanup.

Processed in Siril using mostly the Veralux workflow scripts and a few scripts I made to wrangle JWST data.

I think I'm going to try the level 2 calibrated data next, it'll be more work but I'll also have more flexibility.

I got two mostly clear nights in a row (a rarity this time of year) and was able to gather a total of about 11 hours of sub-frames on NGC 7023. I wish I could get more of the fainter details, but I think I'm limited by the light pollution in my area (Bortle 6ish).
Gear: Dwarf Mini
Settings: eq mode, 60 second exposures, gain 60, astro filter
683 total sub-frames used in the stack

Building on Papillon's original guide on the PixInsight forum, I've put together a consolidated and maintained version on GitLab:

https://gitlab.com/K-22/pixinsight_rocm_linux_instructions

What it covers

The guide walks through building TensorFlow 2.19.1 with ROCm 7.2 support inside a Docker container and configuring PixInsight to use it. The Docker approach keeps the build clean and reproducible — no build dependencies on your host. Also includes performance tuning and troubleshooting.

Tested with

• AMD Radeon RX 9070 XT (RDNA4 / gfx1201)
• Ubuntu 24.04 and CachyOS (Arch-based)
• ROCm 7.2, TensorFlow 2.19.1

Should work with older AMD generations (RDNA3, RDNA2, etc.) by adjusting the target architecture flag — the guide explains how.

Contributions welcome

AMD GPU support on Linux is a moving target. If you've tested a different GPU, found a distro-specific fix, or have a better configuration — open an issue or MR on the repo. Even just confirming it worked on your setup helps.

Full credit to Papillon for the original guide. This repo is just an effort to make it easier to find, maintain, and extend.

Clear skies.

Taken from my Canon EOS T2i, mounted to the top of my telescope. Thirty 10-second exposures. Manually subtracted the same dark frame from every one to account for the many dead pixels, then manually aligned and stacked in GIMP. No idea what I'm doing, but I like it good enough. Stacked with Siril. Should have done that from the start lol

Next time I'm gonna try to capture it high in the sky instead of when it's at the horizon, and I'm gonna watch a tutorial on stacking instead of dicking around blindly

EDIT: I followed a tutorial! At least, as best as I could with one dark frame and no bias or flats. Still, a massive improvement over manual stacking. Still just dicked around with the colors and levels in GIMP after stacking, but you can sort of see M42 now!

Original picture for posterity. This is what the first 10 comments saw.

Equipment:

• Olympus E-M1 Mk III (unmodified)
• OM-System 100-400 f/5-6.3 II
• SA-GTI

Image:

• 30 sec, f/7.1, 1600 ISO
• 245 Lights (~2h)
• Biases, Darks & Flats for Calibration

Stacked using Siril and edited in GIMP

Any advice on how to get more color in?

I've dabbled in amateur astronomy for many years, but I've never really gotten into astrophotography because, frankly, I didn't have the patience. Or the funds. Especially considering where I live. It's a Bortle 6 zone not far from Seatltle, so observing conditions are rarely great. This time of year it's mostly overcast.

In the last several years, though, we've got some amazing new devices coming on the market that make the hobby a lot more accessible. And in the last year or so they've become much more affordable. I just got my hands on a Dwarf Lab Mini. So far, I've only had one good clear night to really put it through its paces, but I'm already blown away at what I was able to capture with no real experience. This is the M81 cluster after about 4.5 hours of 60s subframes. The auto-stacked results looked okay, but I grabbed the FITS files off the device and re-stacked them using Siril, then did some post-processing using Seti Astro Suite. Really not much, though. It almost feels like cheating.

Orions jewel captured Feb 13th

Second attempt at the orion nebula with the dwarf 3.