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

StarNet v2

Astrophotography Imaging Software
  • Please log in to reply
304 replies to this topic

#1 nekitmm

nekitmm

    Vendor - StarNet Software

  • *****
  • Vendors
  • topic starter
  • Posts: 166
  • Joined: 25 Apr 2018

Posted 24 January 2022 - 12:27 AM

Hi, everyone! I am finally releasing an updated version of StarNet - StarNet v2.

 

The link to download is here. Make sure you are downloading the right version.

 

I know that this release is long overdue, but I will touch the reasons behind the delay in the last section of this post.

 

This version took countless hours of work tp prepare more training data, with the push towards star dense images. It was also trained using hardware better than my 10 years old laptop (now deceased...). It should show better performance across the board, including wide field images of the Milky Way (well, also wide field images with no Milky Way in them, if anyone have ever seen those, lol). Hopefully, this will enable StarNet usage for landscape astrophotography. This is ultimately for community to decide, but it works pretty well on images from my rigs. This version should also work well for images at any stage of processing, not only after initial stretch, which makes it more useful.

 

Good news for people who don't have PixInsight - this release includes simple user interface, so you won't need to mess with command line and other scary stuff (hooray!). The negative side of this, though, is that GUI is only available for Windows. I am not able to compile GUI for MacOS, and most likely never will. I know there are ways to run Windows apps on Macs, but don't ask me how, I have no idea. Hopefully the community will be able to help here.

 

This release also includes new version of the PI module. PI team will probably include this into PI distribution at some point, but right now they are busy with other things. This release will install StarNet2 separately and will not replace the original StarNet. The installation procedure is pretty much the same as before and is described in the README.txt file.

 

A few notes on updates and changes:

  1. It was a huge mistake to include a large number of stride values available, and many people wasted a lot of time trying them all out. The new version has only two (256/128) and I highly recommend leaving the default one (256) on!
  2.  Smooth tile transitions will now reduce the number of artefacts.
  3. There is support of linear images (please read on). The neural network itself does not work with linear data, so this feature (if enabled) will use STF parameters to stretch the image first, remove the stars, and then de-stretch the image back. Just make sure to get STF correct before running StarNet. I suspect this will lead to disproportionally large amount of discussion around it with purists arguing how it is not really a linear image anymore after you stretch and de-stretch it back and so on. Still decided to include this, despite people recommending me not to. Please don't make me regret this, don't like it - don't use it.
  4.  There is also 2x up sampling option, which will be explained below.

Known problems:

  1. Standalone version might loose your color profile, so the colors might appear different. If this happens, all you need is to just open the image in the software of choice and restore the profile.
  2. The most annoying problem is that sometimes this version will leave visible patterns around small stars on some images. This tends to affect images with high SNR and tight bright stars. To combat this I recommend up sampling your image (2x). This removes the problem at the cost of higher processing time. There is no option to do this internally in the standalone version, so please upsample your image in PS or other software you use. This problem really bugs me out, but so far I could not get rid of it.

I want to say huge thanks to Rob Pfile, Alberto Ibañez and Eric Coles for their continuing support! This release would be impossible without your help!

 

That's about it for the main part of the post I guess. I hope there will be no problems with PI installation and usage, but the new GUI for Windows might potentially be problematic. I have never built GUI before so I don't know what to expect, but it might not work for some people. Any help will be appreciated.

 

Now a bit on the delay reasons and new license term conditions. Apart from the fact that I started a new position recently and have little time left for my hobbies, including AP, there is another reason behind the delay. This updated version has been sitting on my laptop for a while, and I was not sure how to go about releasing it. I don't know how to describe this in simple terms, but the problem with releasing it is that I have no tools to protect my intellectual property. Conventional software is pretty safe in this regard because once compiled, the original source code is no longer easy to recover. With neural networks based software the situation is different. I spent countless hours manually photoshoping images full of stars (imaging removing stars from wide field image of MW) to get training data. One reason behind this was the fact that there was simply no software out there that could help me (exactly why StarNet was created). However, once such software exists (StarNet), there is nothing that can stop someone else from using StarNet to create training data and train seemingly independent neural network. There are some published research in this area that provides ways to protect intellectual property and I tried to incorporate some of those into StarNet, but I still have no guaranties whatsoever that this will help.

 

Long story short - this release comes with updated license terms that specifically limits StarNet usage to astrophotography processing only. In no form shall StarNet or any of its components and outputs be used to create any derivative works, including but not limited to, training of any kind of neural networks for any king of image processing purposes. This includes not only star removal, but any other kinds of image transformations. StarNet shall not be used to aid development of other software. This limitation includes some non-trivially harmful actions like sharing (PMing) original images on which StarNet works well (and other software does not) with interested parties. Overall, just remember that if something is given to you for free doesn't mean it has no value and you can do whatever you want with it.

 

The open source part of the story is also over, so the code on github will also not be updated, unfortunately. I will also not reply to any inquiries about the details of StarNet's development/training etc. due to reasons I won't disclose now. Any information about StarNet's NN, training/test datasets etc. is now proprietary.

 

Anyway, I really hate writing about these formalities, but I feel that I have to. Above all - just have fun and create cool images!


  • noodle, osbourne one-nil, dswtan and 66 others like this

#2 AarondeVries

AarondeVries

    Explorer 1

  • -----
  • Posts: 86
  • Joined: 02 Aug 2017
  • Loc: The Netherlands

Posted 24 January 2022 - 02:36 AM

Thank you Nikita!

 

You have done a wonderful job. 

I tried it right away with a difficult image (straight out of integration). 

 

HFG1-before.JPG

 

I was never happy with Starnet V1 (on stretched image) in combination with these kind of images from my scope:

 

HFG1-starnet V1.JPG

 

The result with Starnet V2 (on linear image) is so much better:

 

 

HFG1-starnet V2.JPG

 

(Processed version here: https://www.astrobin.com/4tzjyz/0/ )


Edited by AarondeVries, 24 January 2022 - 12:30 PM.

  • Rouzbeh, elmiko, John O'Grady and 10 others like this

#3 R Botero

R Botero

    Soyuz

  • -----
  • Posts: 3,733
  • Joined: 02 Jan 2009
  • Loc: Kent, England

Posted 24 January 2022 - 04:47 AM

Nikita :bow:

Roberto
  • PhotonHunter1 and nekitmm like this

#4 mxcoppell

mxcoppell

    Viking 1

  • *****
  • Posts: 727
  • Joined: 25 Sep 2017
  • Loc: Coppell, Texas

Posted 25 January 2022 - 08:23 AM

Great job, Nikita!

 

Does the new version support the Apple M1 natively? 

 

Thank!

-Min



#5 LucaRavage

LucaRavage

    Lift Off

  • -----
  • Posts: 2
  • Joined: 25 Jan 2022

Posted 25 January 2022 - 10:53 AM

Hi,

 

Very nice, I get much less artefacts on bright stars! (Standalone version on Windows)

 

However personnaly I did like the possibility to use command line, in order to process several files in a batch... Does the .exe of the GUI version accept command line as well?



#6 Readerp

Readerp

    Messenger

  • *****
  • Posts: 421
  • Joined: 22 Jun 2019
  • Loc: So Cal

Posted 25 January 2022 - 11:39 AM

Great!

 

Can you make a GPU enabled version to speed up processing?

 

CS

Pete



#7 nekitmm

nekitmm

    Vendor - StarNet Software

  • *****
  • Vendors
  • topic starter
  • Posts: 166
  • Joined: 25 Apr 2018

Posted 25 January 2022 - 05:14 PM

Great job, Nikita!

 

Does the new version support the Apple M1 natively? 

 

Thank!

-Min

Probably not, I don't have a Mac with M1 to even test it...



#8 nekitmm

nekitmm

    Vendor - StarNet Software

  • *****
  • Vendors
  • topic starter
  • Posts: 166
  • Joined: 25 Apr 2018

Posted 25 January 2022 - 05:15 PM

Hi,

 

Very nice, I get much less artefacts on bright stars! (Standalone version on Windows)

 

However personnaly I did like the possibility to use command line, in order to process several files in a batch... Does the .exe of the GUI version accept command line as well?

No, it won't, I will need to look into this, probably this will need to be a separate version for that. Just out of curiosity, why to you need to process multiple files? Is it just separate channels?



#9 nekitmm

nekitmm

    Vendor - StarNet Software

  • *****
  • Vendors
  • topic starter
  • Posts: 166
  • Joined: 25 Apr 2018

Posted 25 January 2022 - 05:17 PM

Great!

 

Can you make a GPU enabled version to speed up processing?

 

CS

Pete

I imagine most people won't have a GPU, so this might lead to problems. I can probably include separate tensorflow lib with GPU support...


  • RossW and Readerp like this

#10 LucaRavage

LucaRavage

    Lift Off

  • -----
  • Posts: 2
  • Joined: 25 Jan 2022

Posted 25 January 2022 - 07:34 PM

No, it won't, I will need to look into this, probably this will need to be a separate version for that. Just out of curiosity, why to you need to process multiple files? Is it just separate channels?

Yeah, when I process a photo taken with duo-narrowband filter and color camera I first create 4 images in Siril software: Halpha, Oiii, HOO composition and HHO composition, then I make starless versions of each of them at once (hence my use of a batch), and I import all of them in Photoshop as building blocks for further processing. If you can deliver a version allowing batches it would be so great :-) Thanks!


Edited by LucaRavage, 25 January 2022 - 07:34 PM.


#11 nekitmm

nekitmm

    Vendor - StarNet Software

  • *****
  • Vendors
  • topic starter
  • Posts: 166
  • Joined: 25 Apr 2018

Posted 25 January 2022 - 11:38 PM

Update: I was told there was a bug in 2x up-sampling in PI modules. Just updated modules, should be good now!



#12 Jeff2011

Jeff2011

    Soyuz

  • *****
  • Posts: 3,727
  • Joined: 01 Jan 2013
  • Loc: Sugar Land, TX

Posted 26 January 2022 - 09:44 AM

I would like to thank Nikita for allowing me to beta test v2.   My primary usage of StarNet is to remove the stars at a linear stage to get a deeper stretch without bloating the stars and then add the stars back after the stretch using an Adam Block method.   One of my first uses of StarNet v2 was on my Vela SNR IOTD image:  https://www.astrobin.com/p8eb76/0/

 

Thanks a bunch Nikita for all of the effort you put into this and making it freely available.  

 

For those wanting to improve performance, StarNet can be GPU enabled with some additional steps.  

https://www.cloudyni...nabled-starnet/


  • grimlock408, nekitmm, Rich_ and 1 other like this

#13 ChristopherBeere

ChristopherBeere

    Messenger

  • -----
  • Posts: 403
  • Joined: 24 Jul 2019
  • Loc: London UK

Posted 26 January 2022 - 09:47 AM

Can't wait to try it out, absolutely killer module that I use in every workflow.

 

Awesome work Nikita, thank you so much for your dedication.


Edited by ChristopherBeere, 26 January 2022 - 09:48 AM.

  • nekitmm likes this

#14 Dean J.

Dean J.

    Apollo

  • -----
  • Posts: 1,208
  • Joined: 13 Aug 2011
  • Loc: Above the grass.

Posted 26 January 2022 - 04:24 PM

Hi, everyone! I am finally releasing an updated version of StarNet - StarNet v2.

 

Thank you very much for this cool program!  goodjob.gif 

 

 


... Above all - just have fun and create cool images!

I will do my best.  Telescope.gif

 


 


  • nekitmm likes this

#15 R Botero

R Botero

    Soyuz

  • -----
  • Posts: 3,733
  • Joined: 02 Jan 2009
  • Loc: Kent, England

Posted 27 January 2022 - 06:36 AM

Nikita

 

StarNet2 works really well and seems to make use of GPU acceleration if set up as described here:  https://darkskies.sp...t-starnet-cuda/  I used to run StarNet(1) that way and once I installed StarNet2 - without overwriting the tensorflow library from the link above - I got my GPU to provide the memory usage.

 

More importantly, the results are fantastic.  Much, much better than v1 and comparable if not better than some recent commercial software.

 

This is my attempt at MW3 from suburban London:  https://www.astrobin.../full/wwxf0w/D/ for which I made extensive use of StarNet when processing.  The image is 6200x4120 in size.

 

A simple run of v1 on the final image yields:

 

 jwjYbq1qSOLy_620x0_Z-BERGNL.jpg

 

with this Console output:

 

StarNet: Processing view: MW3_LRGB
Writing swap files...
3027.798 MiB/s
Restoring neural network checkpoint: C:/Program Files/PixInsight/library/rgb_starnet_weights.pb
Processing 1617 image tiles: done
56.209 s

 

StarNet2 yields:

 

tYsql0bN-iFc_1824x0_jaUALzhf.jpg

 

StarNet2: Processing view: MW3_LRGB_clone
Writing swap files...
3237.297 MiB/s

_____________              _____   __    _____
___  ___/_  /______ __________  | / /______  /_
______ \_  __/  __ `/_  ___/_   |/ /_  _ \  __/
_____/ // /_ / /_/ /_  /   _  /|  / /  __/ /_ 
_/____/ \__/ \__,_/ /_/    /_/ |_/  \___/\__/ 
_______________________________________________

Starting star removal procedure...
Window size: 512
Stride: 128
Image size: 6200x4120
Number of channels: 3
Color space: RGB
Bits per sample: 8
Has alpha channels: false
Float sample: false
Restoring neural network checkpoint...
Processing 1617 image tiles: done
Done!
01:30.54

 

I am assuming StarNet2 made use of Starnet2_weights.pb in \bin as per installation instructions instead of \library\rgb_starnet_weights.pb as per PxI installation?

 

Roberto


  • elmiko, nekitmm and RossW like this

#16 phtnnz

phtnnz

    Sputnik

  • -----
  • Posts: 27
  • Joined: 11 Oct 2018

Posted 27 January 2022 - 07:58 AM

WOW! What an improvement! Thanks a lot for the work you put into StarNet.

 

I gave it a try with a particularly dense star field around vdB 31.

Results old vs. new version of StarNet.

 

(Now, does this post violate your conditions stated above? ;-)

 

Martin

 

Screenshot 2022-01-27 142211.jpg

Screenshot 2022-01-27 135943.jpg


Edited by phtnnz, 27 January 2022 - 08:28 AM.

  • nekitmm and RossW like this

#17 lucam

lucam

    Apollo

  • *****
  • Posts: 1,213
  • Joined: 09 Oct 2017
  • Loc: Upstate NY, USA

Posted 27 January 2022 - 09:16 AM

Thank you for your efforts and excellent work, Nikita! Also, very interesting considerations regarding derivative works with respect to development of neural networks and AI-enabled software. Definitely food for thought.

 

Best,

 

Luca


  • nekitmm likes this

#18 nekitmm

nekitmm

    Vendor - StarNet Software

  • *****
  • Vendors
  • topic starter
  • Posts: 166
  • Joined: 25 Apr 2018

Posted 27 January 2022 - 07:59 PM

I would like to thank Nikita for allowing me to beta test v2.   My primary usage of StarNet is to remove the stars at a linear stage to get a deeper stretch without bloating the stars and then add the stars back after the stretch using an Adam Block method.   One of my first uses of StarNet v2 was on my Vela SNR IOTD image:  https://www.astrobin.com/p8eb76/0/

 

Thanks a bunch Nikita for all of the effort you put into this and making it freely available.  

 

For those wanting to improve performance, StarNet can be GPU enabled with some additional steps.  

https://www.cloudyni...nabled-starnet/

Thanks a lot for your help, Jeff! Your feedback was definitely useful! Hoping to meet you sometime at the dark site!) It's been a while since my last visit there... frown.gif


  • Jeff2011 and 47FIRE like this

#19 nekitmm

nekitmm

    Vendor - StarNet Software

  • *****
  • Vendors
  • topic starter
  • Posts: 166
  • Joined: 25 Apr 2018

Posted 27 January 2022 - 08:03 PM

Nikita

 

StarNet2 works really well and seems to make use of GPU acceleration if set up as described here:  https://darkskies.sp...t-starnet-cuda/  I used to run StarNet(1) that way and once I installed StarNet2 - without overwriting the tensorflow library from the link above - I got my GPU to provide the memory usage.

Thanks for testing it out! I am glad to hear that the same old trick works! Too bed it is a bit more involved than just copying some libraries....

 

Your result looks great, thanks for sharing! Yes, the libraries are in a different spot than for the built-in PI module, I guess I could move them there for clarity!


  • Jesus Magdalena likes this

#20 nekitmm

nekitmm

    Vendor - StarNet Software

  • *****
  • Vendors
  • topic starter
  • Posts: 166
  • Joined: 25 Apr 2018

Posted 27 January 2022 - 08:06 PM

WOW! What an improvement! Thanks a lot for the work you put into StarNet.

 

I gave it a try with a particularly dense star field around vdB 31.

Results old vs. new version of StarNet.

 

(Now, does this post violate your conditions stated above? ;-)

 

Martin

Lol, no it doesn't!) I definitely don't want to discourage people from sharing their results, the text there describes more specific actions!

 

I am super glad to see you got good results there, the improvement is massive!



#21 qswat72

qswat72

    Viking 1

  • *****
  • Posts: 540
  • Joined: 02 Sep 2020

Posted 27 January 2022 - 10:25 PM

Seems exciting, can’t wait to give it a go! Support for linear images sounds really promising!
  • nekitmm likes this

#22 nekitmm

nekitmm

    Vendor - StarNet Software

  • *****
  • Vendors
  • topic starter
  • Posts: 166
  • Joined: 25 Apr 2018

Posted 27 January 2022 - 11:23 PM

Great!

 

Can you make a GPU enabled version to speed up processing?

 

CS

Pete

As mentioned above, please try this to get it running on GPU: https://darkskies.sp...t-starnet-cuda/

 

Unfortunately it is a bit more involved than just copying files, so there is not a lot that I can do here!



#23 mjrt

mjrt

    Lift Off

  • -----
  • Posts: 2
  • Joined: 27 Jan 2022
  • Loc: Taipei, Taiwan

Posted 28 January 2022 - 12:49 AM

I tested sternet2 and found that it is possible to use GPUs to accelerate computing, but the usage of cuda cores is relatively low, about 60%, compared to starnet, which is 85%.

Attached Thumbnails

  • 272802083_10158987871448402_4280750007650012781_n.jpg


#24 Utkarsh mishra

Utkarsh mishra

    Vendor - Astronomical Image Processing

  • -----
  • Vendors
  • Posts: 6
  • Joined: 13 Sep 2018

Posted 28 January 2022 - 05:21 AM

Hello Nikita 
Thanks for the long awaited release of this! the difference is huge!

Attached Thumbnails

  • starnet2 .jpg

  • nekitmm likes this

#25 rickbassham

rickbassham

    Lift Off

  • *****
  • Posts: 14
  • Joined: 05 Jun 2013

Posted 28 January 2022 - 07:31 AM

Nikita, I don't see a linux build for version 2. Is that coming soon?


  • karolbe and damianc like 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





Also tagged with one or more of these keywords: Astrophotography, Imaging, Software



Cloudy Nights LLC
Cloudy Nights Sponsor: Astronomics