382 replies to this topic

[OldSkyEyes]

Can we see that? I'd much rather focus the discussion around results.

 

Thanks!

Hi,

 

Attached an impression of losing detail between V1 and V2

I do realize that not removing all stars in V1 may give an impression of more detail.

Although some structures are lost while present in V1.

 

Having 2 versions gives the option to chose.

Attached Thumbnails

• 

Edited by OldSkyEyes, 05 February 2022 - 05:29 AM.

[matt_astro_tx]

1. I just uploaded CLI versions for all OS: https://sourceforge....net/files/v2.0/. Look for CLI versions. Let me know if they work lol, I was a bit rushing to get those done.

YES!  *Fist pump*

 

Going to download now.

[matt_astro_tx]

I keep getting this.  Any tips for how to work around it on Mac OS 10.15.7?

Attached Thumbnails

• 

[edif300]

Oh YES !!! What a great improvement . Never liked those star's residual shape in V1but now with V2 is near perfection.

 

Great job Nikita, congrats and thank you.

 

Edited by edif300, 05 February 2022 - 11:08 AM.

[kothemel]

I keep getting this.  Any tips for how to work around it on Mac OS 10.15.7?

You should go to System Preferences -> Security & privacy -> General and allow the file to be executed.

[matt_astro_tx]

You should go to System Preferences -> Security & privacy -> General and allow the file to be executed.

Thanks.  I did a CLI command to do the same thing for every file in the directory, but I'll try it in the preferences screen when my son's done with the computer.

 

EDIT:  The solution was as follows for the Mac CLI version:

1. At the command line, perform the 'chmod +x' command for the starnet++ file, the .sh file, and the two dylib files
2. For example:  chmod +x ./starnet.sh
3. Next, in a finder window, right-click each of the above files and click open
4. In the popup warning that appears for each file, click open
5. You should be good to go

Edited by matt_astro_tx, 06 February 2022 - 05:47 AM.

[dheera]

 

 

1. I just uploaded CLI versions for all OS: https://sourceforge....net/files/v2.0/. Look for CLI versions. Let me know if they work lol, I was a bit rushing to get those done.

Thank you!! The CLI version works, though it only accepts 16-bit images. Is there any way for it to operate on 32- or 64-bit?

[matt_astro_tx]

Got the Mac CLI version working and all I have to say is thank you!  You have saved me countless hours working with the healing brush in GIMP!

 

Here's a comparison on some Rosette data.

 

V1.1

 

V2

[D.Briden]

Help!

 

Trying to get the stand-a-lone version of v2 to work on my MacBook Pro - OSX 10.14.6 Mojave

 

V1 still works fine and I use these easy to follow instructions to copy into the Terminal (I'm not a programmer, so these were uploaded by a SGL member... I'm UK based)

 

----------

5. Open Terminal and navigate to the Starnet folder, copy this command:

cd ~/Desktop/StarNet_MacOS

 

6. Open in Text Editor either the file:

'run_mono_starnet.sh'

 

depending on whether you are working with a mono or RGB file and edit the filenames in there to be the filename of the file that you want to make starless

Save the txt file, then....

7. Drag either 'run_mono_starnet.sh' or 'run_rgb_starnet.sh' onto the Terminal window and you should see the path of the the file appear

----------

 

When V1 is running I get this in the terminal window - as I've just tested:

 

 

So Starnet V2 (in a folder sitting under the V1 version on the desktop named (as downloaded): StarNetv2CLI_MacOS  - with same .tif file as used above....

 

If I double-click the new updated starnet++ application, the Terminal window automatically opens and I get:

 

If I copy the (modified) instruction from 5. above, now with: cd ~/Desktop/StarNetv2CLI_MacOS       - into Terminal and hitting Enter

 

Then open the 'run_starnet.sh' document, edit the file name to that of the mono Jellyfish neb file that I've dropped into the folder, so:

 

#!/bin/bash
./starnet++ mono_jelly.tif

 

I get this:

 

The image is a 16 bit greyscale image (and works in Starnet V1).
Could I ask someone Mac literate to offer a helping hand please as to what the hell I'm doing wrong !!!

 

I have to say, the first version of Starnet++ does a grand job (a little slow - time to go and get a cuppa), but it would be nice to see if V2 does even better and a tad quicker ! 

 

 

 

 

Thanks in advance for any pointers offered....

 

Damian

[_MiR_]

There may be several reasons for this to fail, did you see that there were two demo tif files provided with the download?

 

First step is to go into your starnetv2 directory asnd to check that starnet++ is executable:

ls -l starnet++
-rwxr-xr-x@ 1 ring  staff  474104  1 Feb 22:59 starnet++

if you see only -rw-r--r--@ at the beginning of the output

 

then run

chmod 755 starnet++

and then run:

./starnet++ mono_test.tif test.tif

and check what happens. For me this is the output:

./starnet++ mono_test.tif t.tif
Reading input image... Done!
Bits per sample: 16
Samples per pixel: 1
Height: 712
Width: 1048
Restoring neural network checkpoint... Done!
2022-02-06 13:54:46.170341: I tensorflow/core/platform/cpu_feature_guard.cc:151] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN) to use the following CPU instructions in performance-critical operations:  SSE4.2
To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags.
Total number of tiles: 15
100% finished
Done!

when this works, then just for fun also run the color-test:

./starnet++ rgb_test.tif test.tif

When this step already fails then please provide output you see.

 

Next Step would be to run your image:

./starnet++ mono_jelly.tif mono_jelly_starless.tif

and see what happens....

Edited by _MiR_, 06 February 2022 - 08:08 AM.

[D.Briden]

Thank you _MiR_ for taking the time to reply.

It took me a second to work out your first step.... how to 'go into your starnetv2 directory'- I 'started it' via this command (as for V1 version):

cd ~/Desktop/StarNetv2CLI_MacOS

 

Before that, I also ran this to stop the OSX blocking the software opening:

chmod +x ./

 

for each of these items in the folder:

 

libtensorflow_framework.2.dylib

libtensorflow.2.dylib

run_starnet.sh

starnet++

 

for some reason I missed: starnet2_weights.pb

 

As you expected, I only got: if you see only -rw-r--r--@ at the beginning of the output

 

I then tried to run your suggestion of     ./starnet++ mono_test.tif test.tif.        and yes, I do have the mono and RGB tests in the folder

 

 

I didn't get any further - the result is below:

 

 

Because I got a bit lost and wasn't sure if I'd added the correct file name back into the run_starnet.sh document, I tried again with both the mono and RGB tester images....

 

 

Thanks again for your help!

 

Any more thoughts...?

 

Regards,

Damian

 

[D.Briden]

to double check, I also followed your instructions again, plus made sure that the file name in that     run_starnet.sh      document tied up to the mono test file included. Didn't get anywhere....

[_MiR_]

Two ideas:

one is, did you try clicking on the dylib files as described in #131 ?

 

another issue may be that your os version is to old, and I think this is actually your problem:

 

This command:

dyld_info libtensorflow.2.dylib

libtensorflow.2.dylib [x86_64]:
    -platform:
        platform     minOS      sdk
           macOS     12.0      12.0

shows that minOS of this library is set to 12.0 which is equal to Monteray, your Mojave is 10.14

 

You can try to copy over the libraries libtensorflow.2.dylib and libtensorflow_framework.2.dylib from the Starnet V1 installation, I will try that later but currently the download from sourceforge.net is horribly slow and my copy of old starnet v1 will arrive in an hour or so...

 

unfortunately copying over libs from old version does not work

Edited by _MiR_, 06 February 2022 - 02:10 PM.

[D.Briden]

Hi again _MiR_

 

 

Two ideas:

one is, did you try clicking on the dylib files as described in #131 ?

Yes, tried that - got this:

 

Was going to try copying the files over, but you got there first I see !

 

Is there anything else to try..? 

 

My Mac is set-up for graphics work and I have some software that may not work if I upgrade. Unfortunately, that is more important to me in the grand scheme of things than astrophotography and the need to remove stars... perhaps slightly better than Starnet++ (V1).

 

Whilst I shouldn't criticise the developer for a free piece of software, expecting users to need the very latest OSX to be able to run it is rather.... crazy. Many Mac owners still use their machines for graphics and print work where stability is a must. Major OSX updates have always been buggy, so for those with (in my case) print deadlines - loosing InDesign, QuarkXpress or reprographics software is always a massive risk - hence why we wait and wait for upgrades from both Apple and the other critical software we run.

Would it not be possible to at least have Starnett++ (V2) run from a few older OSX versions..?

 

Damian

Edited by D.Briden, 06 February 2022 - 03:19 PM.

[_MiR_]

The official Tensorflow Libraries

https://www.tensorfl.../install/lang_c

from Google require as a minimum MacOS 10.15, so even Google does not support 10.14 anymore.

 

dyld_info libtensorflow_framework.2.7.0.dylib
libtensorflow_framework.2.7.0.dylib [x86_64]:
    -platform:
        platform     minOS      sdk
           macOS     10.15     10.15

 

also Versions 2.6.0 and 2.5.0 require 10.15

[nwcs]

Assuming your Mac is an intel based one, you’d be better off using VMWare and set up a virtual machine and separate out the Astro stuff. You can have it be on any MacOS that VMWare supports. It would also isolate out potential issues.

[andreya108]

1. I just uploaded CLI versions for all OS: https://sourceforge....net/files/v2.0/. Look for CLI versions. Let me know if they work lol, I was a bit rushing to get those done.

 

3. The bug with linear images and masks in PI appears be solved, but I didn't upload new versions yet. Hoping to get some time to do a bit more testing.

Thank you! CLI version works fine under Ubuntu 21.10, looking forward to PI/Linux version!

Here is the result:

Spoiler

[CptNautilus]

Are you sure you are using stretched images? This is the only potential problem I can think of rn.

  I've found the cause: my TIFF had its data compressed internally (ZIP). It was not mentioned in the file metadata though.

 

  My image is initially in 32 bits and I had to export it in 16-bits and I used the default options, which zipped it.

 

  Looking good! Great app and thanks for sharing it.

[Ripwords]

The cross platform GUI is now updated to use the V2 CLI. To install just download the CLI as you would normally, then download the GUI from this link https://github.com/R...arnetGUI-v0.5.3. Once the GUI is installed, you can then point the GUI to the starnet CLI directory.

 

I've only been able to test the GUI on Windows and Ubuntu VM, so if MacOS users encounter any problems do let me know.

 

Here is a simple demo video of the GUI.

https://youtu.be/OutZCpG2h6E

Attached Thumbnails

• 

Edited by Ripwords, 07 February 2022 - 10:03 AM.

[meegja]

Thanks SO much!! This version 2 makes such a difference!
Below a comparison. Done with the CLI Windows version. Above v1 and below V2. It's the core of the Heart Nebula in Hubble Palette. Especially on the right the difference is huge.

 

I do get the next message though in the CMD window. Don't know what it means or if it's an error? Everything seems to work fine. CPU is an Intel i9-9900kf.

 

Restoring neural network checkpoint... Done!
2022-02-07 17:10:34.108073: I tensorflow/core/platform/cpu_feature_guard.cc:142] This TensorFlow binary is optimized with oneAPI Deep Neural Network Library (oneDNN) to use the following CPU instructions in performance-critical operations:  AVX2
To enable them in other operations, rebuild TensorFlow with the appropriate compiler flags.

 

 

[Ambress]

Hello all,

 

With the update of Starnet Version 2 to also provide a command line executable version, and with a big thank you to nekitmm for having done so, I have now updated a batch script and associated files for launching Starnet Version 2 on the Windows platform in what I favor for my own workflow, and some of you may also like.  Last year I presented this as RGBstarless through a presentation on my YouTube channel.  For the updated version, I have renamed it as Starnet2 Commander, a video for which may be found at https://tinyurl.com/...arnet2Commander and the files downloadable in zip format from https://tinyurl.com/...net2Commander.  This does not modify the files for StarNetv2CLI_Win in any way; it simply changes how the command line executable is invoked, providing access by a drag & drop of files onto shortcut links or by using the Windows context menu "Send to..." option.  If such an approach fits in with your worfkflow style, I hope you find it useful.

Regards,

 

Daryl

Edited by Ambress, 07 February 2022 - 06:21 PM.

[FlankerOneTwo]

Anybody else have issues with Pixinsight crashing after Starnet V2 completes? No error message, Pixinsight just shuts down. I’m using 2x upsampling and 128 stride, with GPU acceleration from a 3080. I kept the original tensorflow file from Starnet 1 to get GPU acceleration working, and I’m wondering if that’s causing my crashes.  

Funny you should mention that, I started another thread before seeing that this one is the preferred mechanism for Starnet support. Don't know if this is the same error you are having, but sounds similar. First off, I think Starnet is brilliant, has really changed my workflow in a very positive way. Here's from my other post (with some added details):
 

I have a weird problem. Starnet v2 has been working very well for me, I typically run it in the linear state, no issues there. I recently acquired data on M81 / M82 in HaRGB - Sn v2 runs mostly fine on the individual channels, but attempting to run on the linear RGB combination results in a crash to the desktop after Starnet completes ("Done!" appears in the Process Viewer), with a Memory Access Violation being logged in the Windows Event logs. No result image is displayed. This occurs 100% of the time with this image. It runs fine on a stretched version of the same image and I can manually relinearize the result without difficulty. Occurs with either the GPU enabled or stock provided versions of tensorflow, so not a CUDA or GPU issue I think. This occurs occasionally and seemingly randomly with the individual RGB channels, but for some reason never with the Ha. I am running with the default stride 256, no upsampling.

 

Faulting module name: PixInsight.exe, version: 1.8.8.12, time stamp: 0x61cce4de

Exception code: 0xc0000005
Fault offset: 0x0000000000287062
Faulting process id: 0x8b50
Faulting application start time: 0x01d81bdd376416bb
Faulting application path: C:\Program Files\PixInsight\bin\PixInsight.exe
Faulting module path: C:\Program Files\PixInsight\bin\PixInsight.exe

 

Windows 10, Ryzen 3950X, 64 GB RAM, RTX 3080Ti, not out of disk/swap space I don't think, I show 113GB free on the System / PixInsight Swap drive. I have paging file System Managed on both C: and D: drives, but max size is only about 74GB anyway. It continues to run fine on the larger RGB linear images I've run it on in the past. Error still occurs when I move Pixinsight swap to a completely different drive.

I have not observed other weird errors in PI or other apps to make me think that this is a CPU / memory failure.

Any idea WTH might be going on??? My guess is that either the relinearization is failing, or an error is occurring during creation or display of the relinearized image, but I have no idea what that might be or why because I can do the steps manually without errors.

Baffled,

Patrick

[polslinux]

CLI works beautifully on openSUSE, thanks a lot mate!

Now waiting for the PI module

[dscottfrey]

on tif files saved out of Starry Landscape Stacker, using StarNetv2CLI_MacOS, I am getting:

libc++abi: terminating with uncaught exception of type std::invalid_argument: stoi: no conversion
Abort trap: 6

tif files saved out of the companion app Starry Sky Stacker, work fine (I am making the big assumption the tif files are output the same). One differences large areas of no stars (a.k.a. the ground) in the SLS tif (I did try cropping out the ground and saving out of photoshop, with the same problem)

Can you point me at anything to solve?Can I send you a test file?

 

[leftyk82]

 

Thanks.  I did a CLI command to do the same thing for every file in the directory, but I'll try it in the preferences screen when my son's done with the computer.

 

EDIT:  The solution was as follows for the Mac CLI version:

1. At the command line, perform the 'chmod +x' command for the starnet++ file, the .sh file, and the two dylib files
2. For example:  chmod +x ./starnet.sh
3. Next, in a finder window, right-click each of the above files and click open
4. In the popup warning that appears for each file, click open
5. You should be good to go

 

I am also on a M1 Mac, without PI.

 

I've gotten mine running currently on 16 stride and it sure is taking its time; seems slower than before but I truly can't recall.  Thirty minutes and only 4% complete.

 

That sound normal?  

 

Edited to add; 7% after an hour

Edited by leftyk82, 08 February 2022 - 06:55 PM.