What's missing from Siril?

[stark ermine]

But as I said. We do it. For each step

Not for a global processing because siril has no idea of what it is doing. Except for step by step.

[trim steppe]

is there no way to implement it without that?

[mental geyser]

No

[stark ermine]

GraXpert is now almost integrated in Siril

[broken peak]

ah that's cool

[trim steppe]

graxpert denoise is pretty good as well ig

[ancient talon]

it's pretty much as good as deepsnr

[mental geyser]

really?

[ancient talon]

• it doesn't hallucinate that many stars and structures

[mental geyser]

are you refering to version 3 denoising?

[ancient talon]

yeah, v3.0.2 or v3.0.1, both are really good

[mental geyser]

right, i need to try those

oh yea its seems pretty good

wow graxpert denoise takes a while

only issue for me is that its not strong enough like deepsnr

[ancient talon]

it's rougly equal to 85-90% deepsnr

[mental geyser]

deepsnr is a noise destroyer while this is more of a "reducer"

[ancient talon]

I mean for meit doesn't matter. I wouldn't use more than like 75% deepsnr anyway, after that it starts to look weird imo

but yeah if you want to really push your data it kinda sucks

[mental geyser]

have you noticed a difference between non cfa drizzled and cfa drizzled data for graxpert denoise?

[raven lodge]

Graxpert and deepsnr do different things. Graxpert does gradients and ai denoise, deepsnr does ai deconvolution.

[mental geyser]

deepsnr is a denoisrre

ai deconvolution is BlurX

[raven lodge]

My bad, I thought it did both. Less excited for DeepSNR then.

Really need a version of BlurX that doesn’t need Pix.

[mental geyser]

theres astrosharp but its quite mid

huh interesting

[ancient talon]

not really but I didn't really compare them

[stark ermine]

Looks like there is a return or a space between - and flat

[mental geyser]

ccompared it with the normal bayerdrizzle script and theres nothing there

they are the exact same

[stark ermine]

there must be a difference, otherwise it would work 🙂

for me, due to the error, you have a return before flat. And flat is taken as a new command

[raven lodge]

Do you mean in terms of the scripts or just in general? I noticed a vast difference between 2x CFA drizzle and non-CFA drizzle with Seestar images in terms of noise smoothness and colour noise, and Graxpert and NoiseX both do a better job afterwards.

[raven lodge]

^

[stark ermine]

I speak about the error in the script

[mental geyser]

So I should remove the flat command?

[stark ermine]

flat is not a command

this is an option of the previous command

you should not have flat on a new line

but on the same line than seqapplyreg

[split smelt]

Seems to work \o/

[compact kraken]

When i try to extract ha and oiii from osc i get the notification
" Not enough star pairs image skipped" in many images and the script fails
I used the l enhance shooting under full moon
The osc preprocessing worked but the osc extract has not.
Can anyone help?

[dense heron]

are you guided ? my guess some stars dont meet what ever roundness threshhold you have. maybe lower that

[compact kraken]

Yes it is guided
How do i reduce the roundness threshold

[mental geyser]

looks like siril just didnt use flats

[stark ermine]

This line is not to correct light with flat. Be careful.

https://siril.readthedocs.io/en/latest/preprocessing/drizzle.html#initial-pixel-weighting

The flat is divided in the usual calibrate command. And in your case you have an overcorrection, probably due to an offset issue

https://siril.readthedocs.io/en/latest/preprocessing/calibration.html#understanding-how-the-flats-correct-the-lights

[mental geyser]

Thanks, will look into that

[split smelt]

are you using sirilic?

if so, go to properties, flat tab and make sure "biases subtract" is ticked

[mental geyser]

Nope. Using script

In siril

[mellow vine]

wish deepsnr was compatible with siril but that's not really a siril problem

[mental geyser]

theres no command line for it yet

if there was, you prolly could

[mental geyser]

when is it expected for a new dev version w/ graxpert denoise?

[stark ermine]

The dev version contains GraXpert interaction now

[mental geyser]

oh

i dont see a thing with all having green ticks

ooh is it one of these

[stark ermine]

I think you are watching old versions

[mental geyser]

oh i dont see it like that

[stark ermine]

Just clicked on the master pipeline

[mental geyser]

those are just keywords

[stark ermine]

?? not on master

you need to download on the master branch

[mental geyser]

im very confused

i dowloaded the windows version and this is it

[raven lodge]

Look in /bin

[trim steppe]

dont think you got the installer

[split smelt]

for Windows, I've found that it's much better to download source code and build it. It's really not hard to do even if you've never done it before. If you want to do it, follow these steps exactly https://siril.readthedocs.io/en/latest/installation/windows.html#building-on-windows-with-msys2

[mental geyser]

gettting this error

when i launch the application

and then it closes

[raven lodge]

I would find the latest build that is tagged siril-win64 that has succeeded to be built and use that. That's what I've been doing.

[mental geyser]

im downloading this version

[split smelt]

I find the Windows builds have all sorts of problems, that's why I build my own.

[stark ermine]

It is better undeed

[stark ermine]

the windows build available from automatic build is built on a Linux system. It can be different from the one we provide at each release.

This is why it is better to compile the code

[split smelt]

looks like something is off in latest trunk. I've got an OIII file that was created by Siril as part of a narrowband OSC stack (HaO3) that I can't open with Siril trunk. 1.2.1 opens it fine. If I save it as TIF, it opens fine in Siril trunk but not as FIT. If I load it into Graxpert, do something and save it, it loads fine in Siril trunk. When I load the bad fit file, Siril console says it's normalizing input data 0,1

[stark ermine]

I have no issue with FITS files on my side

[split smelt]

These two were created from the stacking process. c03 does not open on Siril 1.3.2-alpha. It opens fine on older Siril

[stark ermine]

I can open both images

However, the O3 version has no more signal... Looks like an issue in the calibration step

[split smelt]

open it with 1.2.1 or Graxpert. It's a bit noisy but has plenty of signal

[mental geyser]

Yea we defo need a local version of Gaia catalog

it took me 35minutes for it to download the data from there

my internet is that slow

[split smelt]

@stark ermine I did some debugging. In io/image_format_fits.c, function read_fits_with_convert, case DOUBLE_IMG, fit->keywords.data_max is not initialized so it contains garbage. I guess it was luck that it worked for one of my images but not the other. I copy/pasted the fits_read_key code that you have in the LONG_IMG case and now it works. However, that's not very DRY.

[stark ermine]

This is a double image ??? Where did you produced that?

[split smelt]

Siril produced it via Sirilic

[stark ermine]

Siril doesn't produce 64bits images

[split smelt]

there's no "break" on the case for FLOAT_IMG so I guess both FLOAT_IMG and DOUBLE_IMG is dealt with by the same code

[stark ermine]

Ok. I'm on my phone here.
There are several switch case like that. Will have a look

[split smelt]

if it helps https://gitlab.com/free-astro/siril/-/blob/master/src/io/image_format_fits.c?ref_type=heads#L713

convert_floats gets run in error because fit->keywords.data_max contains garbage

[stark ermine]

I will try to have a look tonight. Because it is a bit more complicated. The statistics computes data_max with a value of 5.387419

This is totally strange this value

[split smelt]

@stark ermine I noticed your commit. It fixes the issue, thanks 🙂

[ashen bough]

Are we gonna get something that improves star processing/ star recomp ?

[ancient talon]

wdym by that? I'm almost done with curves, it's just testing left so I could work on that as my next contribution maybe

[ashen bough]

im asking that weird question cuz i notice whenever i do starnet removal and bring them back with star recomp they always look weird or not as good as they did prior to removing them

idk if thats a skill issue or others notice it maybe 🤷‍♂️

[mental geyser]

ok my brain forgot

hhow do you install the new dev version of siril from the gitlab

[tall sable]

I second this question

[limpid surge]

Make stacking simpler, like one menu that recognizes and stacks the stars instead of one billion menus that are complicated even when following a tutorial

[stark ermine]

Stacking is already very simple. Reducing options is not a good idea.

We don't want an application with only one button.

We want a powerful application

[broken peak]

it looks intimidating at first but it's not that hard to do a basic processing routine manually

[stark ermine]

For very beginners there are scripts: one button to go. Then, curious users can discover all features, doing processing manually.

[limpid surge]

The scripts as well as the more complicated route with a tutorial didn't work for me, it's clearly not 'very simple'

Or are you calling me stupid here?

[broken peak]

what fails exactly? both script and manual way never failed for me unless I did some mistake

[stark ermine]

Not at all, but if you follow official tutorial it is really easy. Did you read it?

Venturing into specialized software without having read the basics is the best way to get disgusted.
And this is not criticism.

[limpid surge]

I did read the tutorial

[stark ermine]

Which one?

[split smelt]

Personally, I find stacking with Siril much easier with help from Sirilic https://siril.org/docs/sirilic/

[stark ermine]

Third possibility indeed

[wooden knoll]

Ahahhaahahahahahaha

No

I hate siril stacking

It was a nightmare that never worked

Meanwhile with DSS it's like 5 clicks and I'm on my way

Siril hates RGGB or smth that's all I could tell from the error messages

Just a terrible experience

But siril is great for like some processing

[stark ermine]

Ok but with DSS you just have basics. Siril is more powerful. Really more

[stark ermine]

Siril hates nothing. Siril is just a software 🙂
He works well with RGGB.

[wooden knoll]

No

Probably something to do with the fact I have a dslr

[broken peak]

I used siril and a dslr for 2 years, no issues

[wooden knoll]

Oh

Ok it just hates me then

[wooden knoll]

What more can it do

[broken peak]

dss only stacks, you can pretty much process from start to finish with siril

[wooden knoll]

Well yeah

I thought we were talking about like stacking features mb

I'll give siril stacking another go

[stark ermine]

I used siril with a DSLR many years

[stark ermine]

The stacking is more sophisticated. With better normalisation . A lot more rejection algorithms, ...

[tall sable]

?

I got the zip from the gitlab repo but I don't know what to do with these files

[stark ermine]

Do you know that now, even some professional astronomers use Siril for it's stacking abilities.

[mental geyser]

I use a dslr and I'm untracked and works great.

[wooden knoll]

Ok

[mental geyser]

Really? Can see why since its incredibly fast since pcs used are quite powerful

If siril isn't stacking your data, then somethings wrong with your data

[stark ermine]

Yes, and this is why we wrote a short article for a scientific journal lately

[tall sable]

did you encounter this when running
meson setup _build --buildtype release
ninja -C _build install

[mental geyser]

Since your here, can you explain how to download the dev versions of siril?

[stark ermine]

I'm not very good with GitLab. I never do it. But you need to find the artifact from the branch you want. However, I always recommend to build the source. It is always better

[mental geyser]

How do you do that

[stark ermine]

https://siril.readthedocs.io/en/latest/installation/windows.html#building-on-windows-with-msys2

but here you find all the artifacts, but for every branch: https://gitlab.com/free-astro/siril/-/pipelines

So it may not be a good idea

[junior hound]

Hey, one issue I have encoutenred is related to debayering : it's really weird so I don't know where it comes from but sometimes, when I debayer the raw files everything works but when I load the sequence, calibrate it and then debayer, the bayer pattern isn't the good one anymore. This wasn't an issue until drizzle integration because I could just change the sequences rgb pattern, but it looks like drizzle integration donesnt take that into account

Anyways the issue is I can't get the pattern right no matter what I try, I have to export it to pipp to flip the sequence vertically and then import it back, so it would be nice at least if you could flip an entire sequence from siril (which I don't think you can do yet)

[stark ermine]

Hum you can update manually your bayer pattern if the one chosen by siril is not good

But generally the choice is good

[ebon wren]

Is siril exclusive to desktop

[trim steppe]

tried running image solver on one of the recent versions and it closes almost immediately after trying, this a known bug?

[stark ermine]

It is exclusive to windows Intel, macos Intel/arm and Linux Intel/arm

[stark ermine]

The dialog closes?

[trim steppe]

siril completely closes

[ebon wren]

Human words pls

[trim steppe]

and now it didnt close, interesting

[stark ermine]

Which version

[trim steppe]

if thats the thing you needed, now it seems to be working though, may have just been a one-off bug on my end, gonna try a few more times

yeah it just works without issues now, dont know what caused whatever happened before

[stark ermine]

Good 🙂

[stark ermine]

What do you think as device?

[ebon wren]

Yes

[stark ermine]

Which device?

[raven lodge]

I’m guessing they mean iOS or Android

[waxen ember]

or perhaps temple os

[tall sable]

is there supposed to be no "output registration" tab in the unstable dev version?

[mellow vine]

@stark ermine are you a siril developer or something

[stark ermine]

Read the red text. Load a sequence first

[tall sable]

i did but it still didnt appear. i figured it out though

[stark ermine]

This tab is being charged currently

Next MR will refactor it

[stark ermine]

@mellow vine : yes

[mellow vine]

is it a paying job

[stark ermine]

No

[mellow vine]

Serious respect for that then

[ashen bough]

Yea zero respects if you were even getting a penny out of this jk

[hasty prism]

I needed to trun pacman -Syu, then run meson setup --wipe _build

after that build with ninja as you would normally

[tall sable]

yeah I did the same and got it working

[mental geyser]

how did you download your dev version i forgor 💀

[stark ermine]

This is how open source project work.
People can make donation if they want, but this is not mandatory. We do it because we love it

[trim steppe]

look for one that has green on all, i got the one with a warning, seems to be fine though, then on the dropdown on the right choose win64 native installer

[stark ermine]

but choose the master branch

other branch could be very instable...

Keep in mind that these versions are not for production work

[mental geyser]

what do you mean by that

[tall sable]

you might want to try creating your own build for it

it takes some figuring out but it’s what I did to get it working

[stark ermine]

What I mean is that this is a development version and we release code almost every day, if not several times a day (except during the summer vacations). It's possible that Siril has bugs, including bugs that can corrupt data. And this is even truer for particular branches. For example, when I reworked the headers of the FITS files, the branch corrupted the FITS files for some time.

[mental geyser]

ah right

[steady mountain]

Personally;

Star Spikes

Allow people in Siril to use a program to artificially add starspikes, or remove starspikes if a user does not want them.

[mental geyser]

Programs like that already exist

[steady mountain]

I know but im too stupid to use the python program

and having it built into siril would make things so much easier

[mental geyser]

You can add diffraction spikes yourself while imaging

[steady mountain]

with a smart telescope?

[mental geyser]

Just use 2 thin rubber bands

[stark ermine]

I will never understand why people want to add artifacts 😢

[split smelt]

Add? Why would you want to do that? Spikes are an image defect

[mental geyser]

Becuzzz it makes you get more likes on binstagram!!!!!!

Star diffractions ruin starless images

[stark ermine]

With the software, we do our utmost to obtain round stars (without coma, which is an optical aberration), as small as possible (by suppressing convolution due to atmosphere and optics). All this to add spikes, which are another optical aberration not visible with refractor. How bizarre. I'm sorry, but I don't want it in Siril.

[steady mountain]

Star Spikes make Stars look a lot more insteresting.

I mean yes if you have extremely round and perfect stars

well done

but if you are not so good at that but want to make your images intruiging to others,

star spikes make stars more interesting, if that makes sense.

[mental geyser]

If you really want star diffractions, just make them

[steady mountain]

though if you could add & remove them that would be useful

[steady mountain]

hard to do with the telescope i use

[mental geyser]

Buy a Newtonian then

[stark ermine]

You can use fishing line.

But again, take it as a chance if you don't have diffraction spikes.

[raven lodge]

It would be amazing if you could execute pySiril scripts from inside Siril, just as if they were normal siril scripts. That would add a huge amount of flexibility, for example a user could create and distribute a script that could react to different combinations of darks or flats, depending on what exists.

If there were hooks to the UI interface Siril uses that would also allow the scripts to pop up dialogues and draw graphs in windows, which I don’t believe is possible with scripts alone right now.

[sage dove]

Artificial spikes look really ugly

If you want spikes make a spike mask

[stark ermine]

But you can

[raven lodge]

You can?

[stark ermine]

Sure

Either you put the script in a folder you load in Siril, or you use the command line with the @ token

https://siril.readthedocs.io/en/latest/Scripts.html

[hasty prism]

not necessarily

[raven lodge]

Wait, you can do that with Python scripts as well as .ssf? All this time I thought you had to call them externally from Python.exe

[hasty prism]

artificial spikes can look really good if done with the right software

I think for some cases, like images that are mostly empty void (e.g. galaxies) spikes make the image look less empty

[stark ermine]

no, I was talking about ssf

[raven lodge]

Ah, ok. I think being able to call Python scripts from inside Siril like you can with ssf scripts would be incredible.

[limber parrot]

whats missing from siril is everything that pix has

[stark ermine]

PixInsight is not a goal to be achieved

why it would be

Siril does a lot of things that PI doesn't do

that's the way life is 🙂

[raven lodge]

This isn't a bug in Siril exactly but maybe a FYI. My process is to stack in Siril, plate solve, SPCC and the normal stuff, then separate the stars using Starnet. The final step on the starless image is Graxpert denoise (external program, not integrated in Siril). Then I pull it back in to Siril using Star Recombination. Very often Siril will flip the starless image along one axis when I do this.

My guess is that when I plate solve, Siril sometimes (correctly) flips the image. Graxpert for whatever reason does not interpret that properly or does not save it properly. So when I read the modified image back in Siril doesn't know it's been flipped.

It happened again to me last night but I can't remember exactly what steps I took. I'll try to pay more attention next time and see if I can reproduce it.

[broken peak]

I usually apply graxpert before removing stars, Idk if it's any better than applying to the starless directly

[raven lodge]

Maybe I should try that. I thought it only worked on fully stretched images though?

[trim steppe]

i do a preliminary rough dbe, then make it starless, port it over to graxpert, get background model, put stars back in and then remove it

or you can just get the starred image and then apply model there

[raven lodge]

Ah, sorry. I'm talking about noise reduction not background reduction. So actually I do pull it in to Graxpert twice, once to do DBE before I remove the stars, back to Siril for a couple of tweaks and star removal, then after stretching the starless image goes back to graxpert for denoising before recombining in Siril. It's that last step which sometimes flips the image.

[trim steppe]

im pretty sure graxpert is meant to run with stars in the image, same way as deepsnr

or the denoise at least

[hasty prism]

both Graxpert AI tools work on linear images

[junior hound]

It actually doesn't seem to work for drizzle integration, no watter what pattern I choose for the sequence (in the settings during import) the exact same one is applied

[mental geyser]

i need to update my siril

[junior hound]

Yeah the new updates are sooo cool

[mental geyser]

soooo how do you update?
do you jus tcopy paste the bin and other folders to siril in program files and replace the ones that are there?

i havent done this

[raven lodge]

It looks like I’ve been using it wrong. I’ll try using it on linear images. I found the -BlurX- NoiseX plugin killed some detail if you used it before stretching and haven’t bothered changing my workflow.

[hasty prism]

blurx has to be used on linear images tho...

[raven lodge]

I meant NoiseX, sorry.

I was using NoiseX with Affinity and it worked well but just as my trial ended the new version of Graxpert came out

[mental geyser]

is this how you do it?

[stark ermine]

If you use development version you can integrate GraXpert into Siril

The way siril read image is bottom up, while other software, like GraXpert use tue top down convention. However, the bottom up is the one that should be used for FITS files

[raven lodge]

Graxpert denoise inside Siril doesn't like my AMD laptop's GPU, but it uses the GPU in the standalone version of Graxpert.

[stark ermine]

Siril just call GraXpert

There's no magic

[stark ermine]

Ask to the GraXpert team maybe

[mental geyser]

I get this error whenever I use ai 3 doe denoising in pixinsight

[tall sable]

https://siril.readthedocs.io/en/latest/installation/windows.html#building-on-windows-with-msys2

[raven lodge]

Out of interest, why should you not select a duoband filter when doing SPCC? The docs say something about it being synthesised in Siril but I don’t understand.

[stark ermine]

@raven lodge : This is because it is too narrow

There is no interest to reproduce the curve experimentally. Same for other narrowband filters

[raven lodge]

What about in the case of wide dualband filters, such as the lp filter in the Seestar? That has 15nm bandpass in Ha and 35nm Oiii. The Oiii is picked up by the G channel as well as B (which I would assume is also true for filters like the L-Extreme when used with most OSC cameras)

[stark ermine]

It's still too narrow for the precision required

[raven lodge]

Ok, interesting. Thanks.

[split smelt]

I always struggle with deconvolution in Siril. Perhaps this is an area that can be improved, either in usability of the feature or in documentation to teach how to use it properly?

[stark ermine]

I would have thought that the documentation was detailed enough. Am I wrong?

There are examples

[split smelt]

I agree, there is a lot of documentation with examples. Despite that, I still can't get it to do anything useful

[hasty prism]

I struggled with decon too in Siril. I tried it again just now and couldn't get it to sharpen the core too much

I crated a PSF image and then extracted stars so I don't have to deal with the ringing

I mean it's doing something, it sharpens the Hii region at the bottom pretty well. But the core is virtually untouched

[trim steppe]

yeah it doesnt do much usually

[junior hound]

Try to use the multiplicative algorithm instead of gradient descent, will do more ringing and add more noise but converge quicker

Also don't create a psf image from the "global registration", manually pick best stars, it's what worked best for me

[hasty prism]

I created the PSF in Pix

[trim steppe]

why would you not use pix decon when you have it

[hasty prism]

to check how it compares to pix's decon

they use the same base algorithm so I figured it'd be a fair comparison. Turns out pix uses different regularization algorithm and that apparently changes a lot

I also wanted to check if people complaining about decon is just skill issue or if it's really just hard to use. For me pix's decon was super easy to use and I got a satisfactory result in ~3min of trying out stuff, no docs or anything, and it was my first time using it. With Siril I guess I just didn't try multiplicative but the above pic is the best I got.

with multiplicative method it worked quite well but also amplified the noise a bit

[junior hound]

Yep thats the main issue... Siril still has some improvements to do! Maybe we'll see some sort of masked decon in the future

[hasty prism]

I honestly don't know much on this subject but I think that's what regularization is for.

[stark ermine]

Big up to @hasty prism who has finished its Curve transformation implementation

It has been merged in Siril dev and will be available in 1.4

[split smelt]

nice! I've been using Gimp for that but this will make things much easier

[trim steppe]

very nice

[sage dove]

Great to see SIril improving

[mental geyser]

[hasty prism]

an advanced all-in-one stacking script is next :)

[split smelt]

To be fair, it improves all the time. The next release is going to be awesome

[mental geyser]

Yooo

[mental geyser]

Tell me more

[hasty prism]

can't promise I'll finish it for 1.4.0 but I'll try

[split smelt]

perhaps just improve Sirilic?

rather than making another script?

[hasty prism]

hm

not sure if an external app is really needed tbh. But yeah just realized it would be "competing" with Sirilic

[split smelt]

the big benefit of sirilic is to handle multiple nights, or batches of different sub lengths. I guess it would be hard to do that with a all-in-one script?

[trim steppe]

holy patchnotes ☠️

[hasty prism]

it would be a GUI tool

[split smelt]

just like Sirilic 😛

[hasty prism]

I planned to add multiple nights, filters and all of the options Siril currently has but to one GUI and to make it all execute on one button

yeah I know. Now that you point that out I'm not so sure I'll add that anymore lol

[split smelt]

perhaps you could merge your ideas into Sirilic?

[raven lodge]

Sirilic should be part of Siril IMO

[stark ermine]

Sirilic is in Python

this is a different project

[hasty prism]

I'd have to learn python. But then again I never touched C before Siril so maybe

I'll talk to the dev, see what I can do

[raven lodge]

The functionality, I mean, not necessarily merging the projects. Sirilic provides stacking functionality that should really be part of the core app.

[trim steppe]

does it have cfa/bayer drizzle?

[hasty prism]

idk, probably not

[stark ermine]

I think the project is a bit behinf Siril

[hasty prism]

new drizzle is on the dev version and I'd imagine the guy behind Sirilic only updates it after Siril updates

[stark ermine]

now we have 4 devs working on Siril. He's alone with Sirilic

Sirilic is sync with the stable version, no problem with that

But the dev one,.... THis is a huge step forward

[split smelt]

No. You have to build the script & then edit it. It's only a small change though

I usually tick "2 pass", then build siril script, then edit it. Remove all keyword "-drizzle" then find the seqapplyreg line and add "-drizzle -scale=1.0 -pixfrac=1.0 -kernel=square"

that's it

[stark ermine]

btw, you can contact the dev of Sirilic for asking him to update the dev version interaction. He's very active when users ask something (when he reads the message on the gitlab)

[split smelt]

it's probably a moving target for him right now

[stark ermine]

probably yes

even for me this is difficult to follow everything 🙂

too many contributions, too many commits a day

a big refactoring of the registration is coming. With distorsion management

[hasty prism]

oh wait really?

I wanted to suggest that as an alternative contribution lol

[stark ermine]

😉

[hasty prism]

What are you using? Thin plate splines?

[stark ermine]

no. Using the WCS with the platesolving

[split smelt]

perhaps you can get _krzy to finish the mosaic stuff? 🙂

[stark ermine]

mosaic is on its way I believe

same dev working on it

distorsion is an important step to get mosaic working

1.4 will have mosaic

fore sure

This is the big feature remaining

[hasty prism]

I'll find something. Or I'll start working on a landscape MW stacking app. I wanted to do that at some point

[split smelt]

I did a mosaic with Siril recently. I used astrometric registraton in 1.3 for alignment but then resorted to letting MS ICE do the actual mosaic combining

[hasty prism]

Sequator is meh, it doesn't handle distortions well and lacks some features the dev refuses to add

[stark ermine]

Implementing an equivalent to sequator would be great

yes. The stacking of Siril is better, adding the equivalent of sequator in Siril would be definitely game changer in star landscape stacker

I believe we have a ticket for that ^^

Oh yeah... THis one, a French one, submitted by a user: https://gitlab.com/free-astro/siril/-/issues/259

I need to translate the issue

Anyway. Off to bed

cheers

[split smelt]

It's ok. Chrome does that

@hasty prism should I be able to move the black point in the curves transformation dialog?

[hasty prism]

First and last points are only moved vertically

That's by design. Do you think it should be changed?

[split smelt]

I can add a 2nd point on the lower horizontal and then that becomes my new blackpoint

it would be better if you could move the first point horizontally

[hasty prism]

That shouldn't be hard to implement. I don't remember why I decided against that when I was writing it but I can't find a good reason

[split smelt]

I just thought of a few other features that would be great in Siril...

ability to constrain some of the transforms to a selection e.g. saturation, colour correction, even stretching. curves, etc. Also add circle selection, perhaps even freehand, or some intelligent selection. And the ability to feather the selection

also add a simple "unpurple" filter to remove chromatic aberration from stars

[hasty prism]

or masks

it's super easy. Probably one of the most requested features in Siril

[split smelt]

either a mask or selection or whatever, just to allow existing transformations to apply to only part of the image

the other thing I'd like to see. An option to record your session to make GUI operations be saved to a text file as scripting commands to make a repeatable history of your actions. Of course, this assumes that every GUI thing is also doable as a scripting cmd but I think that's mostly the case anyway.

[junior hound]

Yay!

Once again, love to the siril team

As soon as I'm getting an income I'm donating as much as possible lol

[junior hound]

Regularization sort of do that indeed but not the same way a mask would. It has pros and cons

Honestly I never got egularization to work well in Siril

However I did played around with the decon and found that using a few multiplicative iterations and an adapted stopping criterion, it works quite well without adding noise

[chilly marten]

Has anyone here used siril to plot a exoplanetary transit? I know that it sort of has the capability’s

[limber parrot]

erm i need to update siril god damn

ooo siril in microsoft store

finally my stupid anti virus wont think siril is trying to dox me

[stark ermine]

https://siril.org/tutorials/photometry/

[stark ermine]

Sadly, the version in the Microsoft store is too old

[limber parrot]

it turned out i couldnt even download it from the microsoft store anyways so i just disabled my anti virus for the time of downloading

[stark ermine]

This is the thing to do. AV are just a pain in the ass

[mental geyser]

the antivirus is the virus

[stark ermine]

About the masking feature. This is scheduled for 1.6

[mental geyser]

lol

whats 1.5 then?

[split smelt]

I would guess the next alpha version after 1.4

[stark ermine]

Indeed

Odd number are for dev versions

[mental geyser]

didnt know that

i eed to download the 1.4 dev version

except ive got no idea lol

[stark ermine]

Explanation here:
https://siril.readthedocs.io/en/stable/Issues.html#send-us-useful-information

This is the gimp convention

[eternal inlet]

Does it apply to Star Recomposition?

[stark ermine]

No. It must be applied on already stretched images in general

[chilly marten]

Why is the background extraction the only tool in siril that requires you to apply before closing it

[sour canopy]

Ai deconvolution. (Blurex)

[trim steppe]

wait until vikram cooks up the alternative to that

[mental geyser]

thats been delayed

hes had personal commitments

[hasty prism]

graxpert has an active deconvolution branch on their repo

maybe they're doing a NN based decon?

oh yeah it seems they are

[trim steppe]

will be interesting to see how that turns out

becoming one hell of a software package

[split smelt]

@stark ermine would a purple defringe filter be accepted into Siril if I coded it?

[stark ermine]

• Ctrl+I
• SCNR
• Ctrl+I

🙂

[split smelt]

yeh I've never had much luck with that. It turns it from purple to cyan

I usually use the "unpurple" filter in Gimp on the starmask layer

[stark ermine]

oh ok

I don't know this filter

[split smelt]

it's part of g'mic plugin https://gmic.eu/

[stark ermine]

yeah I know G'MIC, again a French dev 😉

[eternal inlet]

Because it allows you to apply multiple iterations to smoothen it out better.

[hasty prism]

It's not

[split smelt]

some early results. Far from finished yet

[trim steppe]

bwoah

[split smelt]

this is blue channel, before and after

[trim steppe]

how it even do that

[split smelt]

it's nothing magic - just the same stuff you can do in Gimp. Detects purple using a hsv range, applies a luminance mask using green channel, then reduces the blue content by an amount

loving this ROI thing in Siril btw. I didn't even know it existed until earlier this week

[limber parrot]

thats really cool

1 star seems to appear out of nowhere

[split smelt]

yeh it's there though, not invented, just hidden by the purple fringing. To be honest, I overstretched the stars in that image just so I could test this. I'd usually not get that much purple 🙂

[stark ermine]

Did you code this in Siril ??

@split smelt ?

[split smelt]

yes

it didn't take long

[stark ermine]

Oh great. Do you want to open a MR in Siril ?

[split smelt]

it was fun to do even if it's not something you want to include

[stark ermine]

You contact DM me

It could be a nice filter

[split smelt]

ok. I'll fiddle with it some more until I'm happy with it

[stark ermine]

Ok. DM me for help. I could have a look tonight (on my TZ)

[visual roost]

Hello, I'd like to report a bug:
Photometric color cali. for M51 does not work. I've tried it on 3 different images and each time the stars could not be matched. It works fine on other objects though

[stark ermine]

It is not necessary a bug

It is probably not a bug

More information needed to understand

[chilly marten]

Try platesloving the image and then put those exact coordinates in

[stark ermine]

We need logs. Is it astrometry that fails or photometry... Etc...

[chilly marten]

Btw @stark ermine the photometry thingy in siril doesn’t work so great. It seems like the comparison stars don’t add anything to the plot. Other than being able to check them for anomaly’s

[stark ermine]

This is not the case

You can even check the code

Why do you think this?

[chilly marten]

I’ll post a comparison between siril and AIJ in a minute

[stark ermine]

You are not showing the light curve here

So this is normal

[chilly marten]

How would I do that?

[stark ermine]

These are just individual plots

You have a light curve button

Please do not hesitate to read our tutorials

And in the dev version, we have reworked a lot this area

[stark ermine]

https://siril.org/tutorials/photometry/

[chilly marten]

I need to get gnu plot first, that didn’t work on my Mac. I’ll try on my windows laptop later

[stark ermine]

Oh ok. If you can use the dev version then. You don't need gnuplot

[stark ermine]

@chilly marten : and I really recommend to use dev version for photometry

[chilly marten]

Where can I download that?

[stark ermine]

Either you compile the source, our documentation explains that, or you find binary on our repository

[chilly marten]

Alright, I’ll have to look when I have time for that. When will the next update release?

[stark ermine]

When it's ready. No idea

[split smelt]

@hasty prism can we make it so the curves transformation window can be resized like the other transformation windows?

[ancient talon]

It was originally. I was told that it's better to not make them resizable because there were some issues with that. You need to ask @stark ermine for more details

I think some dialogs might still be resizable because their code wasnt touched in a while

[stark ermine]

Indeed. Normally the dialogs can't be resized because of some issues, especially on macOS

[split smelt]

ok, but histogram & GHS can be resized

[junior hound]

Oh ok, but I agree that it would be nice to have the ability to resize them somehow, at some point in the future. It can get hard to make the precise adjustments at the default scale

[raven lodge]

The star recombination panel likewise. In advanced mode you have to scroll to get all the GHS options

This is why everyone uses Electron. Not because it’s good but because all the cross-platform gui libraries suck.

[split smelt]

what is Electron?

[junior hound]

@stark ermine I have a small list of little life improvements I think would be useful and easy to implement, how can I send it to the dev team?

[raven lodge]

Framework for apps that is essentially just a Chromium window rendering a local web app. VSCode, Slack, etc.

[split smelt]

ok but Siril is not a web app, it's an application written in C

[raven lodge]

I’m sure it can be compiled to webassembly 😛

Also you can write your app in C and just get it to handle the gui. But it was more a comment about how all gui toolkits suck rather than suggesting rewriting it in Electron.

[stark ermine]

here is a good start

[stark ermine]

GTK is fine. The problem is the lack of macOS dev

GTK4 exists but .... the upgrade will take time

[split smelt]

are you using GTK3 right now?

[stark ermine]

yes

[split smelt]

is anyone working on a GTK4 upgrade?

[stark ermine]

no

We tried, but too much work

we need to work on other things

[stark ermine]

So here this is a good idea, but it's important to bear in mind that, at present, almost all the features have been implemented. The stabilization phase will begin shortly.

[junior hound]

Perfect, I'll do tonight

[tall sable]

is it possible to add pause stacking mid way?

it slows down my computer and sometimes I want to play games but my fps is like 15

[mental geyser]

dont play games and stack at the same time

siril stacking doesnt take very long so just hold out lol

[stark ermine]

not easily.

[stark ermine]

Don't do that. Siril compute free memory before the stacking. If you play during stacking you will have memory issues.

[floral plaza]

Idk if its been mentioned yet, but maybe add a way to allow the pre-processing script to run even if calibration files are missing?
I tried preprocessing without flat frames and it didnt allow me, so i had to do it manually :pp

[mental geyser]

https://gitlab.com/free-astro/siril-scripts/-/tree/main/preprocessing?ref_type=heads

[floral plaza]

ahhhhh alr tytyty

[sage glade]

Uhm how can align the images @stark ermine? I have taken the data form https://discord.com/channels/794642864218439681/1281571807438311476 in order to do edit practice and his narrowband stacked image is slightly tilted compared to his RGB and I can't make them alinged :|Thumbnails attached

[stark ermine]

in 1.2.x version, it is better to create a temp sequence and to align images with the classical global registration algorithm.
However, in the dev version the registration is automatic is one in one click in this module

[sage glade]

Yeah but the sequence doesn't involve stacking? Or can I do it with these 6 channels alone?

[stark ermine]

sequence does not involve anything no.
You select all the file, convert them (ideally with symbolic links ON!!) then you register. Once done you can rename each registred file as you want

[sage glade]

Ok thanks I will try that and see 🙂

[sage glade]

Ok umh @stark ermine if you can help me a little.I learn how to do sequence how to align extract and then use RGB to make and image.But a friend of mine said that I can use pixel math to make better results and gave me some equations can I do something like that?

[trim steppe]

pretty sure thats for adding narrowband to an rgb image in pixinsight lol

[sage glade]

Yup and he said that pixel math is similar in siril so I could do it too

I don't think that's the case but agree to disagree

[trim steppe]

there are differences for sure, the star addition thingie thats commonly used didnt work on siril

[sage glade]

Bruh there are really no videos in pixel math which I understand what the other one is doing

😦

[stark ermine]

THis is equation for continuum subtraction

$T does not exist in Siril, you need to update equations

[sage glade]

Ok thanks @stark ermine it'll be awhile before how I learn how to do that but I will do it 😄

[stark ermine]

split your RGB into 3 channels. Then load the 3 channels into the pixelmath. Replace each $T by the adequat channel

[sage glade]

Ok thanks I will try tomorrow

[ancient talon]

just in case, you need to perform continuum subtraction first. This is for adding CSed narrowband to broadband data

just simply adding NB data to BB data can sometimes ruin the colors

[stark ermine]

Continuum subtraction is very useful to enhance very faint signal

Here, a test I've done on my data (negative view)

[trim steppe]

the actual oiii arc

[stark ermine]

indeed

[fallen glen]

How much data is that

[stark ermine]

18h of OIII. The weather is not good 😦 I need to add more data

[limber parrot]

bro if you can get 2x more data the oiii arc is gonna be insane

[stark ermine]

Yes. Will be patient

And of course, 100% with Siril 😉

[mental geyser]

[dense heron]

i tried getting the developer build since im told it has real drizzle

i followed steps on free astro

[tall sable]

1 sec

ChatGPT to the rescue

[broken peak]

...

[dense heron]

chat gpt better hook me up

wait should i uninstall my original siril ?

[split smelt]

if you follow this literally to the letter, it will work fine https://siril.readthedocs.io/en/latest/installation/windows.html#building-on-windows-with-msys2

[dense heron]

i did that one first and rn did the one on free astro as another attemp since its slightly differnt but no

i straight up copy paste it all

in order

the blue msys goober

[stark ermine]

Yes. This I the only one to follow. The free astro one is deprecated.

[stark ermine]

I think you forgot 🙂

[raven lodge]

You don’t need to build on windows if you just want to use it. Just download one of the generated builds.

[limber parrot]

@stark ermine since i dont usually use siril is there 1x drizzle?

[stark ermine]

in the dev version there is CFA Drizzle yes

[limber parrot]

oh wow

how much can you drizzle?

like what range

for example 1x - 3x

[mental geyser]

yes

[limber parrot]

sooo 1x, 2x, 3x?

because that was just an example

[mental geyser]

no

[limber parrot]

bruh

[stark ermine]

you can imagine not integer too

[limber parrot]

woah

so for example 1.5x drizzle is possible?

i might start to use siril for stacking tbh

[stark ermine]

it will be possible yes

[junior hound]

Yeah it's great, besides the spreed an variety of algorithms the drizzle is a really good option

The better noise profile completely changed my pics

[limber parrot]

how does siril stacking compare to pixinsight's?

in speed and results

anddd the storage it takes

mostly the storage

[junior hound]

Indeed! Sorry it's my first week in college so I forgot

[junior hound]

I havnt used pix's stacking but from what I've heard it's faster

Storage idk but it should be about the same

[limber parrot]

ah but whats faster exactly? siril or pix

[junior hound]

Tho you have a lot of control on the registration - stacking in siril

[junior hound]

Oopsie I wrote faster than I thought, siril is faster

Not sure if it's still true when using drizzling

[limber parrot]

aight

thanks for answering my questions

i think i will switch to siril idk yet

i mean in stacking since i dont really know how to process well in siril

[junior hound]

Compared to the videos I've seen of people using pix, siril looks less cumbersome but I can't guarantee

[tall sable]

I love how fast siril stacks things

[junior hound]

Understandable! The work flow in siril is quite different so don't mind asking me or others if you have any questions

[raven lodge]

Stacking with CFA drizzle is a lot slower than without but it’s still fast. I registered & stacked 1500 Seestar frames (1920x1080) the other day with 2x drizzle and it took less than 30 minutes.

[broken peak]

Stick to pix if you have it, Imo while it takes longer the drizzle comes out better

tbf I haven't managed to replicate pix drizzle settings in siril, I mean the .8 dropshrink and varshape 1.5

[trim steppe]

is this in a version you have to build yourself?

[limber parrot]

ahh alright thanks for letting me know

[stark ermine]

I don't think there are a lot of differences between both

[broken peak]

do you know what kind of settings I have to use to match pix's?

if I could open the xsif file for my stack I would compare the two drizzles

[stark ermine]

This is not how we should think. We shouldn't try to find parameters to look like Pix, but parameters to get the best out of the algorithm used :).

I really can't tell about PI 🙂

[broken peak]

so I should try with siril's default drizzle parameters?

[stark ermine]

Not necessarily. You should try with different parameters to find what's the best for your images

[broken peak]

I'll have to try yeah, I've got some new data and pix trial expired so I guess it's time for siril to shine

[junior hound]

I just use the point drizzle, it works best for me

Perfect for denoising

Noise profile is great

Sure varshape might be slightly less noisy/more adapted when less data is available, but it's essentially just blurring a point drizzle

I never stacked in pix so I cant compare them tho...

[broken peak]

I'll try that next

tbh I have no idea how much changes between the methods, I use varshape beacuse that's what deepsnr needs

[trim steppe]

yall built the version youre using rn yourselves?

[broken peak]

nope, mine was already built but it's also not the most recent version

downloaded from this page, I think it was a win64 folder
https://gitlab.com/free-astro/siril/-/artifacts

[trim steppe]

seems im a bit slow with the builds that are floating around then

[junior hound]

I regularly check for "fully compiled pipelines", that include the win64 installer

It's only available for 1 week after being built

[broken peak]

on the same page I linked or somewhere else? the win64 I downloaded now doesn't launch

[junior hound]

Ha I've never used the link you sent so I can't tell

[trim steppe]

pipelines page

[junior hound]

https://gitlab.com/free-astro/siril/-/pipelines this is where I download from, but there's no win installer available rn

This is the last build I have

It should work just fine

[trim steppe]

downloading this one rn with a warning

[broken peak]

installing it now

[junior hound]

Where did you find it? I must have skipped it lol

[trim steppe]

page 2 on pipelines page

[junior hound]

Ok so I'm blind basically

[junior hound]

You might as well download the same build as our dear concord, it's more recent

[broken peak]

yeah doing that now

[raven lodge]

Varshape is like a modified Gaussian, I saw during a 2am deep dive a few weeks ago. Has a slightly flatter “peak”

[broken peak]

2am deep dives hit hard

[junior hound]

Btw the point drizzle is what worked best with dsnr for me

[trim steppe]

was gonna say the same thing

[junior hound]

Same for graxpert's denoise

[broken peak]

there's a dsnr version that works with siril?

[trim steppe]

any artifacting whatsoever?

[junior hound]

It was my pix trial 😭

[broken peak]

ah lol

[junior hound]

Tho I'm pretty sure he plans to do a standalone version like starnet one day or another

[broken peak]

yeah I heard that too, like a year ago sigh

[junior hound]

Graxpert showed a few artefacts and just a touch of false stars (not surprising given how noisy my data was) but not a single false star with deepsnr

[trim steppe]

he gotta get himself some help so it goes faster

[junior hound]

Yeah I would gladly help him

I've already made an app to visualize astro images and edit them, I could offer him to implement deepsnr in it

[trim steppe]

pretty sure he needs it more for making the cli and actually training the stuff

[junior hound]

Ye can't help with that

except if I continue my own AI but I don't have the time now that I'm in college

There's vikram and his decon as well, wonder where he's at

[trim steppe]

where do you find this tab, i might be a tad lost

[broken peak]

registration, apply existing registration

you have to first register with two pass, then apply

iirc

[junior hound]

Yep exactly

[trim steppe]

noted, dont think i have the space for that though

[junior hound]

2 pass global alignement
Apply existing registration and "use drizzle"

[broken peak]

I don't think it creates new files?

like another batch of r_light

[junior hound]

2 pass doesn't require 2 times more space if that's what youre worried about, it only stores registration data

[broken peak]

and you can always delete the older files in the process folder

[junior hound]

The new file is only created once you apply the registration

[trim steppe]

well it takes like over 100gb to stack all the data anyway since it creates the intermediate files during conversion

[broken peak]

that's a lot of very small drives

[trim steppe]

im broke

[junior hound]

What kind of hellish storage system is that

[broken peak]

understandable

[trim steppe]

i have 1 more hdd but i cba to try and slot in the sata cable behind the mess that is my pc

the H drive is also old asf and sounds like a geiger counter

[broken peak]

mmh radiations

my oldest drive is from 2012 I think

[stark ermine]

Accurate photometry is important for the PCC and SPCC color calibration processes, therefore if these are to be used later in the workflow it is strongly recommended to choose a drizzle kernel that preserves flux.

[trim steppe]

who cares about accurate colours when you can have pretty colours

[broken peak]

oh yeah I forgot about that, was it only square that kept it?

[stark ermine]

All users who like to use SPCC

[junior hound]

Pretty sure point does too

[broken peak]

mmh I'll do both and see if I notice any difference

[junior hound]

Nice

[broken peak]

mh applying registration just fails

[trim steppe]

im guessing you cant run the cfa drizzle stuff on master stacks?

[junior hound]

Wdym?

[broken peak]

it needs the subs

[trim steppe]

yeah thats what i thought, no way to weasel between doing all new stacks ig

[broken peak]

unfortunately no

[stark ermine]

You are doing a CFA drizzle on already debayerd images

Can't work

[broken peak]

damnit, I forgot about that

[stark ermine]

Do 'ot hesitate to have a look to the documentation.

[broken peak]

it's alright I'm just an idiot, I did the process a bunch of times already

not enough apparently

[broken peak]

welp, weight by fwhm didn't even try to remove the ice spots that formed over the sensor

[stark ermine]

Looks like hot pixels

Bad dark subtraction?

[broken peak]

I'm pretty sure it's the ice that usually forms, I have to leave cooling running for a while before they're gone

removed the weighting

welp, gaia is timing out

help? I think something went wrong

this is after registering with the point model

[trim steppe]

dsnr works on it?

[trim steppe]

holy lines batman

[broken peak]

don't have pix anymore to try

[trim steppe]

you moved to siril fully?

[broken peak]

pix trial expired

@junior hound point gave me a much worse result, Idk if I have to change something to make it better

[ancient talon]

really cool! I was curious why most software limits drizzle to an integer scaling

[broken peak]

weird things happening here, not sure what's up with the colors

[stark ermine]

How many subs?

[broken peak]

480?

it's 23h or so

[stark ermine]

ok. Is this the stack result? or just one registred frame?

[broken peak]

stack

did the usual conversion, calibration, two pass and point thingy

the stack with generalised extreme [...]

[stark ermine]

what scale?

pixel fraction?

[broken peak]

all left to default

[junior hound]

Wow

Ok seems like a complicated start

[stark ermine]

1 and 1 ?

[junior hound]

Gaia is timing out but I think it's a probleme with the database not siril (I have the same issue)

[broken peak]

yes

[broken peak]

ye I know, it worked a bit later

[junior hound]

I suppose you dither enough ?

[broken peak]

every 3 subs by 15px

should be enough over almost 500 subs

[junior hound]

Not sub pixel dithering?

[broken peak]

Idk I just set that value in nina

it's what I've always used

[junior hound]

Well I'm not experienced with dither (I have poor tracking so it naturally dithers well) but I think you need subpixel offset to have best results

[broken peak]

I mean it doesn't dither by exactly 15px, I assume whatever I'm using is enough

[junior hound]

And unfortunately if your offset is integer I'm pretty sure point won't work

[junior hound]

Yeah but might not be enough

[broken peak]

wait are you talking about acquisition or processing?

[junior hound]

Acquisition

[broken peak]

where can you set sub pixel dithering?

[split smelt]

If you share your subs, I can try to stack them for you to see if it's a local issue with your copy of Siril?

[junior hound]

I have no idea sorry, I'm neither a Nina user or a dithering expert

But I've heard people talk about it so I guess it's a thing

[split smelt]

You can't

[junior hound]

That could be good indeed

[broken peak]

then I remembered correctly, I never heard of sub pixel dithering

I can try to upload the subs but they're a lot

[junior hound]

Ha I'm surprised

[split smelt]

no tracking is 100% accurate, so any dither will be "subpixel"

[junior hound]

Cause drizzling doesn't really care about exact pixel dithering

It needs every location inside the pixel to be centered by some frame

[split smelt]

I challenge you to even get 2 subs perfectly aligned on a pixel boundary 🙂

[junior hound]

Gosh I don't know how to explain it

[junior hound]

Depends on the focal lenght, my frames don't move by one pixel in like 10 minutes when I'm 80mm

Ok I'll think about it and try to come up with a visual explanation

But waiting for that @broken peak have you tried using square or turbo model? (turbo is just square but faster when there's no rotation involved)

[broken peak]

the other pics I uploaded were with square

#1220109001682059275 message
#1220109001682059275 message

[stark ermine]

The Gaia archive status indicator is not responding. This does not necessarily mean the Gaia archive is offline

[junior hound]

Have you used hot/cold pixel removal?

And dark optimization?

[broken peak]

dark optimization no but is it really needed on a 533?

hot cold pixel removal I can't remember

[split smelt]

Mine neither, when guiding. But after you've told the mount to dither, do you think the 2nd image will start perfectly aligned with the pixels of the first image? Even the best mount will be slightly off hence your "subpixel" dither for each sub

[junior hound]

It's not, I was checking if you disabled both (with hot pixel removal) because they tend to mess up and introduce artefacts like what you experienced

[junior hound]

I see what you mean but there's a problem with that, I'll try to make a drawing to explain

[split smelt]

I know we're looking to have an offline database for spcc, but I also think we should find a way to host an online database. We could make it a community effort where people host a mirror & we have some mechanism where we load balance across the mirrors & can gracefully handle failures

if you're interested, I could put a few ideas together

[stark ermine]

I think the priority is offline 🙂

[junior hound]

@broken peak Sorry I never had those issues, I don't know what to advise you...

[broken peak]

I think I'll wait to gather some more data and try to get another pix trial

[junior hound]

It would be nice if figured it out with siril first

https://jwst-docs.stsci.edu/jwst-near-infrared-camera/nircam-operations/nircam-dithers-and-mosaics/nircam-subpixel-dithers/nircam-standard-subpixel-dithers

So subpixel dithering is indeed a thing for professional telescopes, it works like I thought it would

I'll try to make a drawing

[broken peak]

I'm not sure what to try tho

[broken peak]

definitely not something for amateurs like us lol

[junior hound]

If you still have the uncalibrated frames have you tried unchecking cosmetic correction before calibration?

[junior hound]

Ye but I think we can learn from that

[broken peak]

I'll try tomorrow

[junior hound]

Cause my dithering is better as it goes through all pixel positions (at the cost of bad tracking), and yours is not as good because well you are guided and I'm poor
Consequence of that is you have to used some sort of gaussian kernel to make sure all pixels are fully covered, when I can use the point model which will make for the best denoising profile possible and give ever so slightly sharper results

[junior hound]

Ok, I'm going to sleep soon too

[broken peak]

trying to do a quick stack with only darks and no flats as it would take longer

[junior hound]

Yep and you can also try to use similarity transformation if you don't have too much distorsion

A bit faster, a bit less wonky at times

[stark ermine]

There's no reason why it shouldn't work with Siril. We've integrated the HST drizzle, so it's effective and it works.

[broken peak]

it more to see if there's really any difference with how pix does stuff, I had some older data with the same ice issue and I got no artifacts

[junior hound]

I don't think the artefacts from the ice issue are related to drizzling

Only issue I see with drizzling is not perfect dithering which makes the point model too "selective"

[broken peak]

I've dithered and drizzled the same with my iris nebula, 380 images or so and I didn't get any kind of artefact

[junior hound]

Only difference with pix, drizzling side, would be the kernel model since they offer a varshape model (which I'm pretty sure is just a moffat distribution, might be implemented in siril)

[junior hound]

From what I understand but can't say to you clearly because I'm too tired , the issue only manifests with the point model which is more selective

On your result with the square model, the image is looking quite nice overall except for the area with ice

[broken peak]

the colors get really weird after removing stars, I have no idea why

[stark ermine]

Gaia is back

[junior hound]

Your channels are linked in the autostretch preview?

[broken peak]

can't remember tbh, I've closed siril now

[mental geyser]

why does this download 1.2.3

version of siril

[junior hound]

1.2 is still being updated so be careful, you should download the artefacts from the master repository

[junior hound]

The 1.2 repository is written under the title btw

[mental geyser]

[broken peak]

looks better now, I just used winsorized sigma clipping to stack

[stark ermine]

Siril 1.2.4 has just been released

with a fix for macOS user and available disk space computation

[mental geyser]

[dense heron]

For the Siril drizzling, with the version that has real drizzle. I can script it or no ?

Or is the dev version different

[stark ermine]

All Siril version are scriptable

[sage glade]

Hey @stark ermine can you show me how to do a continium subtraction in Siril?

[mental geyser]

pixelmath

[sage glade]

A video or Pictures with arrows I don't understand another way 😦 I tried

[stark ermine]

Yes this is with Pixel Math, I update this tutorial:

https://www.nightphotons.com/guides/advanced-narrowband-combination

[mental geyser]

its jimmys tutorial!

[stark ermine]

And it's a good tuto. Too bad it's not finished

[mental geyser]

hes over on r/ap discord server, couple of us have been hampering him to add the faint addition

hes also written a bunch of other guides like MSGR, single channel deepsnr, independant starless processing

[stark ermine]

Yes. ANd an update with Siril would be awesome 😄

[split smelt]

that might be a good addendum to Siril documentation - how to adapt Pix pixelmath formula to Siril

[ancient talon]

To be fair faint NB addition is basically just HDRing your continuum subtracted image

[mental geyser]

theres no iHDR in siril

[ancient talon]

He said he'll be writing HDR as the next tutorial. Expect it in a year or so

[mental geyser]

he said hes gonna make a broadband processing guide but no idea when thats coming out lol

[stark ermine]

i would like that community start to work on it. Not always me 🙂

Tutorials must come from outside too

[ancient talon]

HDR in Siril would be easy if there was an implementation of more robust multiscale processing

Wavelets are meh for this use

[stark ermine]

Someone wrote this script: https://gitlab.com/free-astro/siril-scripts/-/merge_requests/13

from three linear images taken with different exposures.```

[split smelt]

I'll try to get an article on my website that we can either reference or adopt

[mental geyser]

looks interesting

[ancient talon]

This is for stacking multiple images to do HDR.
I meant HDR stretching/DR compression. GHS is alright for that when the dynamic range isn't too big but for very bright objects it makes bright features very desaturated

[split smelt]

@stark ermine are there plans to improve the live stacking feature so it doesn't create so many additional files? or at least to clean up afterwards?

[stark ermine]

The Live stacking is absolutely not at our top priority yet. But if someone want to work on it, it will be with pleasure.

We start to think what will be 1.6 ... Maybe we will work on the planetary algorithms

[waxen ember]

or maybe landscape 👍

[broken peak]

what's the right way to merge Ha with osc? I've extracted the Ha, aligned it and added it with the rgb compositing tool but I'm not sure it's the right way to do it

[ancient talon]

Jimmy's CS tutorial is probably the way to go. It's 100% reproducable in Siril

[broken peak]

the advanced narrowband combination tutorial?

I had something from him saved, I think it's the right one

[ancient talon]

yep, that's the one

[broken peak]

nice, I'll look into it then

lol graxpert completely nuked the Ha in my first attempt

[mental geyser]

yea dont use graxpert

[ancient talon]

I use graxpert to denoise an image prior to CS and never had it eat singal

That looks more like star removal did a poor job. Graxpert is awesome for denosing before continuum subtraction because it doesn't nuke bright stuff like deepsnr does

[mental geyser]

i often need to run graxpert denoise 2 times since one isnt enough

still lots of noise

[ancient talon]

damn for real?

it's roughly equivalent to 0.85-0.9 deepsnr

[broken peak]

I need to check that, I didn't think about starnet eating stuff

[mental geyser]

well, ive got low SNR data so it really does test denoising

[broken peak]

I haven't been able to do CS btw, I think my Ha data isn't good enough

[ancient talon]

mind showing a screenshot?

[broken peak]

I can send one tomorrow, I was heading to sleep

[waxen ember]

https://open.spotify.com/artist/0NznkrZSzt0dasLdsjCDku?si=KcSqxgjVTGiIfedLvgIbwQ

this you @stark ermine ?

[mental geyser]

lol

[raven lodge]

This is pretty vague... anyone have any idea?

[ancient talon]

Have you tried checking the log?

It's probably a corrupted file if I'd have to guess

[raven lodge]

That is the log 😄

[ancient talon]

Isn't there a log.txt or something like that?

[raven lodge]

Can't seem to find one in the workspace or siril directories. It might be somewhere else but windows file search is terrible.

Clicking the "Export the log" button just exports what's in the console.

This is stacking, so it should only be reading files that siril itself has created in the registration phase :/

[ancient talon]

Check if one of the files it wrote isn't corrupted

[raven lodge]

There's 1500 images...

[waxen ember]

maybe try stacking in something else like dss

if it goes through that might mean it's a siril issue

[ancient talon]

Try only stacking the registered files in DSS

[raven lodge]

Just deleted all the working files and tried again. Same result. In the console (the windows console, not the one in Siril, I see this -
allocating data for 12 threads (each 589 MB)
log: Starting stacking...
partial read from FITS file has been requested outside image bounds or with invalid size
partial read from FITS file has been requested outside image bounds or with invalid size
partial read from FITS file has been requested outside image bounds or with invalid size
partial read from FITS file has been requested outside image bounds or with invalid size
partial read from FITS file has been requested outside image bounds or with invalid size
partial read from FITS file has been requested outside image bounds or with invalid size
partial read from FITS file has been requested outside image bounds or with invalid size
log: Error reading one of the image areas
log: Pixel rejection in channel #0: 0.000% - 0.000%
log: Pixel rejection in channel #1: 0.000% - 0.000%
log: Pixel rejection in channel #2: 0.000% - 0.000%
free and close (0)

I'm using 2x drizzle with the maximum field of view

[raven lodge]

DSS complains that there's a image width mismatch, which there would be because it's using maximum stacking

[broken peak]

Are you sure there really isn't an image with a slightly different resolution?

Try to register in dss and sort by resolution, it should be apparent if there's something wrong

[raven lodge]

The originals are all 1080x1920. Siril’s choking on the drizzle and/or maximum framing somewhere. I’m just trying to stack them with normal framing, see if that works.

[stark ermine]

funny. We have seen this error in a bug report. But we cant reproduce

LOoks like a memory error

https://gitlab.com/free-astro/siril/-/issues/1366

We only see it on macOS. Did you use macOS?

[raven lodge]

Windows 11. I don’t think I was out of memory or SSD space but I wasn’t paying attention to those figures.

SSD has plenty.

BTW the stack works fine just using normal framing.

[waxen ember]

what's your hardware specs?

[stark ermine]

When I speak about memory, I speak about RAM

[stark ermine]

ok. The registration changed a lot in the dev version

maybe it will be different

[mellow vine]

@stark ermine please make xisf files usable in siril

[stark ermine]

https://siril.readthedocs.io/en/latest/file-formats/XISF.html

[mellow vine]

Ok I get that but like

Could I alteast drop my xisf into siril and get a question asking if I want it to get converted to fits

other software like graxpert can use xisf

[broken peak]

can't you just open a xisf and save it as fit manually?

[stark ermine]

Siril dev can use XISF as explained in the link I sent

In read only mode

So Siril will be able to convert XISF in FITS, to read XISF, but not to save into XISF file.

[waxen ember]

why is that exactly? can't you convert it the same way graxpert?

[stark ermine]

and it is already implemented in the dev version

[stark ermine]

it is explained in the link

We don't want

[waxen ember]

okay, thanks for clearing it up

[raven lodge]

I am using 1.3.2-alpha-dev commit 325071e, a few weeks old

[stark ermine]

Yes, but there is a special branch for all the changes in registration 🙂

It will be merged once all feature are done

[raven lodge]

On a different topic, what is the syntax for specifying a filter name in spcc? The docs say that the names of the params must match those in the dialog box exactly, but those have spaced in them. I've tried 'spcc -oscsensor=ZWO Seetar S50', 'spcc -oscsensor="ZWO SeeStar S50"', 'spcc -oscsensor=ZWO_Seestar_S50' and they all have syntax errors. the last one is the .json filename from the repository. OK, got it.

BTW I think the ASI462 and Seestar S50 sensors are mixed up in the repository. Not a big deal since they're the same sensor, but still...

[trim steppe]

i had an interesting bug before that seemed reproduceable, loaded images into dev version, did colour cali and stuff, then ran starnet, starmask came out all weird, had like the background of the starred image and then got a prompt when trying to run histogram stretch on starless to rescale the image since the pixel values were borked

[stark ermine]

The prompt is normal

About the bug, I'm not sure I understand. Are you sure you check the linear checkbutton?

[trim steppe]

yeah, tried on stretched and linear with correct settings, same thing

[stark ermine]

We need more information like screenshots, original file to reproduce, etc....

[trim steppe]

give me a moment, just got home

[trim steppe]

forgor to take screenshots of artifacts

[stark ermine]

This prompt is new, but this is normal

this is not a bug

[broken peak]

what's the right thing yo do in that case? I've use the rescale all so far

[stark ermine]

That depends what you want ^^

[split smelt]

I've seen that prompt after loading a TIF saved by Gimp and rescale gave me real good results once I reset the black point

[broken peak]

that's good to know

[tall sable]

same here

[trim steppe]

wasnt only the prompt, the images got a bit borked as well

[stark ermine]

Why did you used DSS?
You could have some issue with the stack of DSS....

Also you have a terrible gradient

[trim steppe]

i did the starnet stuff in pix and it worked fine

gradient is relatively simple to get out with dbe as well

[stark ermine]

If starnet works outside, it works in Siril too. Siril just call Starnet

[stark ermine]

No. not deeply

[mental geyser]

saying you use DSS to a siril dev is like stabbing them

[stark ermine]

not really 🙂

But DSS is doing some strange things to linear data...

[mental geyser]

it does interesting stuff with the master bias

[stark ermine]

Anyway, StarNet works fine in Siril too

even if the gradient is always here

[stark ermine]

master-bias is the frame not really needed 😉

[trim steppe]

what does the generated star mask look like?

[mental geyser]

its used in the calibration process isnt it?

[stark ermine]

The generated star mask is just a single operation: Original - starless

[stark ermine]

Can be replaced by synthetic biases... I always use it in Siril

[mental geyser]

oh i just have a library of bias frames on my laptop

[stark ermine]

https://siril.readthedocs.io/en/latest/preprocessing/calibration.html#bias

and : https://siril.org/tutorials/synthetic-biases/

[mental geyser]

didnt you guys have a section on wy you shouldnt use dark flats?

[stark ermine]

yes we have

https://siril.org/2021/12/enough-with-dark-flats/

[mental geyser]

oh its on the siril website...i was trying to find it on the docs

[stark ermine]

I know this text caused quite a stir on the networks. Yet the biggest names in astrophotography in France (Christian Buil, Thierry Legault) are doing what we do...

Anyway, for me StarNet works like it does. But I would recommend to use Siril for stacking. It is really better.... no joke

[trim steppe]

maybe i chose the wrong option on the rescale, idk

dont have the space to stack in siril though

[stark ermine]

Don't use scripts then

You would have more control

[trim steppe]

might give it a try but i have like 70gb free at most

[broken peak]

That's not too bad, you just have to delete the previous step every time

[split smelt]

I purchased a used 480GB SSD for ~£25. If you only have USB3 connector rather than SATA, then a USB3>SATA converter is less than £10 on ebay. Storage is cheap these days

[trim steppe]

i got 10eur on my account and i need a dew heater

[split smelt]

I had that recently. Both of my cheap usb dew heaters broke almost at the same time, so I built one of these. It's fantastic https://sourceforge.net/projects/arduinonanodewcontrollerpro/

[sage glade]

Hey @stark ermine do you have any plans for siril to export into .avif or .webp sometime in the future?

[stark ermine]

Already done in the dev version

[sage glade]

https://tenor.com/view/surprised-sorprendido-shaquille-oneal-gif-23222312

Bruh I can't believe I can have a 2 mb picture with 95% of the quality of a .tiff file

[stark ermine]

well, webp we may have removed something because it was bugged

[sage glade]

these are insane formats

[stark ermine]

but for sure we can handle jpgxl wich is very good

[sage glade]

.avif?

Actually scratch that .avif is not supported by discord 😦

[stark ermine]

Yes, okay. Actually no, we had intended to put them in and didn't, in favor of jpgxl I think. I can't really remember because it dates back to the beginning of the dev version.

[sage glade]

Bruh. Is it really better to have a jpgxl or not? I haven't tried that format

[stark ermine]

It can handle lossless compression for 32bits images

Yes, it is good

[sage glade]

https://tenor.com/view/vince-mcmahon-intrigued-gif-27408103

[stark ermine]

I see we have avif too

[sage glade]

Damn and is this going to be at the 1.4.3?

[stark ermine]

Oh, no, I'm sorry. I was right. SaveAVIF has been removed from Siril. I can't remember the reason

[stark ermine]

jpgxl yes

1.4.0

[sage glade]

I'm exited I wanne see how it'll 😄

[stark ermine]

AVIF and HEIF read support for reading ICC profiles is complete.
AVIF / HEIF support for reading files with NCLX profiles is partially complete, but the most likely cases are covered and I think it's complete enough for this MR (it's as complete as gimp).
Documentation MR is updated.```

[sage glade]

https://tenor.com/view/homelander-based-the-boys-homelander-the-boys-facts-gif-26206051

That will actually improve the expirience massively

Hey and how what kind of file sizes are you talking about @stark ermine ?

For example a 3008x3008 png is at 42-47 mb

whereas the jpg is at 18mb

[waxen ember]

jpg is lossy so you trade some data for smaller sizes

[sage glade]

Yeah but he's talking about jpgxl

[waxen ember]

it will be closer in file size to jpg than png and have better quality than jpg

it's also the only* common image format that supports 32-bit

[stark ermine]

so the perfect format for deep sky images storage.

[sage glade]

Amazing thanks @stark ermine

[sage glade]

Hey @stark ermine do you have any idea as to why when I'm in autostrech the image appears like that?

and when I stretch it from the GHS or histrogram is like that?

[stark ermine]

What is the issue?

[sage glade]

In the image in autostrech Siril shows it as a painting with hard corners whereas when I stretch it manually it shows it a photo

[sage glade]

There

[stark ermine]

You need to use HD Autostretch I guess

sometimes, 8bit is not enough for rendering