Jump to content

  •  

CNers have asked about a donation box for Cloudy Nights over the years, so here you go. Donation is not required by any means, so please enjoy your stay.

Photo

siril-cli is MUCH MUCH faster than Pixinsight WBPP on Mac M1

  • Please log in to reply
17 replies to this topic

#1 etotman

etotman

    Lift Off

  • *****
  • topic starter
  • Posts: 9
  • Joined: 01 Mar 2021

Posted 31 October 2024 - 02:53 PM

Typically Pixinsight WBPP takes 30 minutes to an hour on my Mac M1.  For fun I tried a simple siril-cli script to fully process 297 images from last night, and the entire process was done in 6 minutes!  I removed the stars, because that's the way I like it.  PSA for the minority of us with very long integration times in Pixinsight on a Mac.  I welcome any tips since this is my first siril script.  Well not really mine, since I copied snippets from all over.

#!/bin/zsh

initdir=$(pwd)

######## Set your own variables #############
SCRIPTS_DIRECTORY=$initdir
SIRIL_PATH=/Applications/SiriL.app/Contents/Resources/bin/siril-cli
#############################################

# Removing process folder if exists #
rm -rf $SCRIPTS_DIRECTORY/process

echo "Running siril script in $initdir"
$SIRIL_PATH -d $SCRIPTS_DIRECTORY -s - <<ENDSIRIL
requires 1.0.0
convert a3 -debayer -ser -out=process
cd process
seqsubsky a3 -rbf
register bkg_a3 -drizzle
stack r_bkg_a3 rej 3 3 -norm=addscale -output_norm -out=../result

close
ENDSIRIL

echo done

chmod u+x script.sh

script.sh *.fit

Attached Thumbnails

  • LBN 673 nebula_small.png

  • AstroPhotog likes this

#2 etotman

etotman

    Lift Off

  • *****
  • topic starter
  • Posts: 9
  • Joined: 01 Mar 2021

Posted 31 October 2024 - 02:57 PM

LBN673, Rasa 8, ASI533mc, 120s x 297.  No flats, darks, etc.



#3 bbasiaga

bbasiaga

    Skylab

  • *****
  • Posts: 4,420
  • Joined: 10 May 2006

Posted 31 October 2024 - 03:38 PM

I've never used PI.  But I use SIRIL with its OSC-preprocessing script all the time.  Same thing.  Takes about 6-10min.  I usually have a couple hundred subs, plus the calibration files.   

 

I've heard PI WBPP is 'better' but Siril is definitely faster!   And I like the results.  Maybe you can test some out and tell us if you really see a difference between a Siril stack and an a WBPP stack of the same data. 

 

I would not do the test with drizzle though.  From what I understand Siril's drizzle algorithm is not true drizzling.  But you'll want to look in to that as I'm only regurgitating what I recall reading elsewhere.  

 

Brian


  • mariemarie and etotman like this

#4 vidrazor

vidrazor

    Fly Me to the Moon

  • *****
  • Posts: 6,082
  • Joined: 31 Oct 2017
  • Loc: North Bergen, NJ

Posted 31 October 2024 - 04:17 PM

Nico Carver did a stacking showdown with a multitude of programs a while back, and P.I. and A.P..P. were the absolute worst apps for stacking.

 

Even if you do all you image processing in either of those apps, just stack in Siril and save yourself the headache.



#5 nwcs

nwcs

    Apollo

  • -----
  • Posts: 1,120
  • Joined: 12 Nov 2013
  • Loc: Tennessee

Posted 31 October 2024 - 04:40 PM

Use FastIntegration in PixInsight either with the option in WBPP or FBPP. That’s the recommendation for more than about 80 images. Orders of magnitude faster.
  • Zambiadarkskies likes this

#6 klangwolke

klangwolke

    Mariner 2

  • *****
  • Posts: 218
  • Joined: 24 Dec 2023

Posted 31 October 2024 - 05:06 PM

I wish I had a better understanding of what quality you’re going to lose in a typical multi night scenario by using FB vs WB vs Siril.

#7 nwcs

nwcs

    Apollo

  • -----
  • Posts: 1,120
  • Joined: 12 Nov 2013
  • Loc: Tennessee

Posted 31 October 2024 - 05:52 PM

I wish I had a better understanding of what quality you’re going to lose in a typical multi night scenario by using FB vs WB vs Siril.

FWIW I personally use FastIntegrarion in WBPP and get the best of both of those and I’m almost always in a multi-night scenario.



#8 bbasiaga

bbasiaga

    Skylab

  • *****
  • Posts: 4,420
  • Joined: 10 May 2006

Posted 31 October 2024 - 07:38 PM

I wish I had a better understanding of what quality you’re going to lose in a typical multi night scenario by using FB vs WB vs Siril.

That Nico Carver video mentioned will give you some examples. IIRC, the answer is 'not much'



#9 klangwolke

klangwolke

    Mariner 2

  • *****
  • Posts: 218
  • Joined: 24 Dec 2023

Posted 31 October 2024 - 07:51 PM

Thanks, I’m going to check the video out.

#10 Zambiadarkskies

Zambiadarkskies

    Mercury-Atlas

  • -----
  • Posts: 2,550
  • Joined: 17 Nov 2021
  • Loc: Zambia

Posted 01 November 2024 - 01:02 AM

FWIW I personally use FastIntegrarion in WBPP and get the best of both of those and I’m almost always in a multi-night scenario.

Same here.  I have done my own comparisons between PI WBPP and fast integration and APP.  I could not see any difference at all.  I used to use APP becuase it is so much faster than WBPP, now I just use fast integration.  The speed difference is huge.  


  • nwcs likes this

#11 etotman

etotman

    Lift Off

  • *****
  • topic starter
  • Posts: 9
  • Joined: 01 Mar 2021

Posted 01 November 2024 - 08:24 AM

The other reason I now prefer siril-cli is the fact that I can process multiple imaging sessions quickly and easily from the bash prompt, no GUI needed.  I think Pixinsight has a command line interface too, although I haven't tried it.



#12 etotman

etotman

    Lift Off

  • *****
  • topic starter
  • Posts: 9
  • Joined: 01 Mar 2021

Posted 01 November 2024 - 09:40 AM

Actually Pixinsight with fastintegration enabled was a minute faster than siril-cli, but I prefer the siril result.  Unprocessed integrations below.  With some tweaking in Pixinsight I may be able to improve the result.
 

Attached Thumbnails

  • lbn673-integration-test.png

Edited by etotman, 01 November 2024 - 09:44 AM.


#13 etotman

etotman

    Lift Off

  • *****
  • topic starter
  • Posts: 9
  • Joined: 01 Mar 2021

Posted 01 November 2024 - 10:18 AM

This time I enabled WBPP max quality, fast integration.  This increased integration time to 11 minutes.  siril-cli is faster, and the quality is better.  This is a beginner forum, and I admit that I'm a Pixinsight beginner, so I could be missing something important.

Attached Thumbnails

  • Screenshot 2024-11-01 at 8.08.08 AM.jpg


#14 t-ara-fan

t-ara-fan

    Skylab

  • -----
  • Posts: 4,181
  • Joined: 20 Sep 2017
  • Loc: 50° 13' N

Posted 01 November 2024 - 12:50 PM

Are we comparing apples and apples?  Does Siril do Local Normalization?  The LN step in Pixinsight wbpp takes at least half of the time.


  • dswtan, terry59 and Zambiadarkskies like this

#15 DeepSky Di

DeepSky Di

    Fly Me to the Moon

  • *****
  • Moderators
  • Posts: 5,130
  • Joined: 15 Aug 2020

Posted 02 November 2024 - 03:42 PM

Nico Carver did a stacking showdown with a multitude of programs a while back, and P.I. and A.P..P. were the absolute worst apps for stacking.

 

Even if you do all you image processing in either of those apps, just stack in Siril and save yourself the headache.

I watched this again last night. PI didn't come out worst, but it did take 20 minutes where Siril took 2 and a bit. APP in the video took 34 minutes, but 

 

In the video which is 2 years old, Nico didn't know about the APP short cut to go straight to tab 6 and integrate without doing anything on the other tabs. When I do that, APP takes about 2x the Siril time which would have been 4-5 minutes and second fastest if done that way in the video. Doing this makes APP use its defaults which are generally good, and explained with hover text. For example in a recent project I did not have to do anything special to incorporate several subs with aircraft crossing them.

 

Also, APP does not save intermediate files when finished. It writes some temporary .dat files during stacking which get deleted at the end.

 

Before stacking in PI it's necessary to know what to stack, i.e. blink and remove bad subs.

 

In APP I now jump to tab 3 first, which creates the quality graph. I spend some time kicking out bad subs and then continue to tab 6 to integrate.

 

APP handles mutli session as well as multi filter, and also mosaics. I just did an integration with files from last year and this year as separate sessions. A session is essentially a set of files that need different calibration files.

 

Altogether I have preferred stacking in APP - it's fast enough and way faster than WBPP. 



#16 unimatrix0

unimatrix0

    Fly Me to the Moon

  • *****
  • Posts: 7,062
  • Joined: 03 Jan 2021

Posted 02 November 2024 - 03:49 PM

I watched this again last night. PI didn't come out worst, but it did take 20 minutes where Siril took 2 and a bit. APP in the video took 34 minutes, but 

 

In the video which is 2 years old, Nico didn't know about the APP short cut to go straight to tab 6 and integrate without doing anything on the other tabs. When I do that, APP takes about 2x the Siril time which would have been 4-5 minutes and second fastest if done that way in the video. Doing this makes APP use its defaults which are generally good, and explained with hover text. For example in a recent project I did not have to do anything special to incorporate several subs with aircraft crossing them.

 

Also, APP does not save intermediate files when finished. It writes some temporary .dat files during stacking which get deleted at the end.

 

Before stacking in PI it's necessary to know what to stack, i.e. blink and remove bad subs.

 

In APP I now jump to tab 3 first, which creates the quality graph. I spend some time kicking out bad subs and then continue to tab 6 to integrate.

 

APP handles mutli session as well as multi filter, and also mosaics. I just did an integration with files from last year and this year as separate sessions. A session is essentially a set of files that need different calibration files.

 

Altogether I have preferred stacking in APP - it's fast enough and way faster than WBPP. 

As I told in another topic, PI starts struggling when there are a lot of bad subs, with star trails or very noisy images. 

I use SubframeSelector tool, which helps to batch-quality check the subs based on the star shape/size noise level and a whole lot more other criteria to sort out the subs and get rid of the bad ones.   It's far far better to use than blink, because blink depends on just our eyesight while subframe selector measures star sizes and background noise and many other things and it's a lot faster to pick out the bad ones from an interactive graph. 

ASTAP does similar, but less sophisticated measurements. 


Edited by unimatrix0, 02 November 2024 - 03:51 PM.

  • etotman likes this

#17 etotman

etotman

    Lift Off

  • *****
  • topic starter
  • Posts: 9
  • Joined: 01 Mar 2021

Posted 03 November 2024 - 11:22 AM

As I told in another topic, PI starts struggling when there are a lot of bad subs, with star trails or very noisy images. 

I use SubframeSelector tool, which helps to batch-quality check the subs based on the star shape/size noise level and a whole lot more other criteria to sort out the subs and get rid of the bad ones.   It's far far better to use than blink, because blink depends on just our eyesight while subframe selector measures star sizes and background noise and many other things and it's a lot faster to pick out the bad ones from an interactive graph. 

ASTAP does similar, but less sophisticated measurements. 

Good tip!  I enjoy using PI.  It has many useful tools and scripts.  I needed to reprocess several years of images quickly and easily.  siril-cli does the job without any tweaking, and it's fast.  I like the results, and I can easily batch process many image sets.



#18 Oort Cloud

Oort Cloud

    Fly Me to the Moon

  • -----
  • Posts: 6,922
  • Joined: 19 Nov 2020
  • Loc: New Jersey, USA

Posted 03 November 2024 - 01:59 PM

That Nico Carver video mentioned will give you some examples. IIRC, the answer is 'not much'


Guaranteed, it won't be anywhere even close to the difference in time. Siril is so fast. I process/stack manually, so it's one step at a time. It's so fast that sometimes I'll hit the button to create a sequence, or calibrate it, and will think the click didn't register because it finished in less than 1s.
  • etotman likes this


CNers have asked about a donation box for Cloudy Nights over the years, so here you go. Donation is not required by any means, so please enjoy your stay.


Recent Topics






Cloudy Nights LLC
Cloudy Nights Sponsor: Astronomics