I missed that the average star FWHM is show in arc seconds, not pixels which GraXpert requires. When you click the Average PSF button, look in the console and you'll see a line that looks like this: Found 15012 Gaussian profile stars in image, channel #1 (FWHM 4.420661) That FWHM is in pixels, and is what you should use in GraXpert. Sorry about that...
There's nothing to apologize for and it should be transferred automatically in the next version 0I0 In addition, this value that appears in the console was not documented anywhere and was probably incorrectly labeled FWHN, which shouldn't be a problem. Deleting the focal length doesn't really make sense to me because it only changes the value for FWHMx/y Super support😁👍
It looks a little different in Siril 1.2.5 when I click average PSF: (What value do I use besides the middle of FWHMx/y?) Here's what is showing: Average Gaussian PSF N: 91 stars (3 saturated and excluded) B: 626.016567 A: 5586.769714 FWHMx: 6.58" FWHMy: 5.95" r: 0.905 Angle: 0.14 deg rmse: 1.191e+02
Sorry, what looks different? Per my comment above you need to use the value show in the console not in that Average Star Data. Or am I misunderstanding?
Really cool. Siril 1.4, new features, and GraXpert with new features, all integrated. The future is bright ✨. Very informative, thanks very much from (cloudy 😬) Germany 👍🔭
Ahh , does make a difference. I didn't notice at first, but I'd forgotten to check my fwhm average in siril. Once I input that correctly into graxpert, there was a marked difference, even with dwarf data
@@JLCubing the main downside of pix is that it is designed to look like software from the 90s, its incredibly slow and tedious for simple things (especially layer based editing/why do we have to have a separate preview window for EVERYTHING?) it can be faster if you learn it well yes but its still designed badly
@@TheOfficialProtostar that's just pixinsight's design language. and the part when you mentioned slow; you need a good pc for it and allat for it to run fast. I dont mind with the extra windows cuz I don't do real heavy processing lol
@JLCubing no when i said slow i meant its slow to operate because of its design not because "you need a good pc" haha, i have a 13th gen i9 and a 4080 my pc isnt slow, and im familiar with many complex softwares
@JLCubing and the extra windows i mean, why do we need to have a "preview window" to show us what a simple damn curve is doing? 99% of photo/visual software overlays it ON your photo and has a way to disable the preview. Even with ways to do layer based merging in pixi its so backwards, why do i need to install a llugin or script to be able to easily use blending modes that also take several seconds to update instead of instantly lmao
Wow! Great feature! One question: with Deconvolution now available in GraXpert, what would be the "proper" workflow? Should I run Crop, Background Extraction, and Deconvolution in GraXpert, then move to Siril for PCC, StarNet removal, and GHS to adjust the background image with proper saturation? After that, should I return to GraXpert for denoising (linear data) and finish with star recomposition in Siril (or Photoshop)? What do you think of this workflow? Before GraXpert's Deconvolution feature, I used to run Siril's Deconvolution right after PCC.
I use GraXpert to run BGE, Decon, and then Denoise. Then jump in Siril for PCC, star removal, stretching, and then add back the stars. That's just my process though. FYI, all 3 functions in GraXpert should be ran while the data is still linear, not stretched. You can do it after stretching if you want though. It's just the recommendation. In the end, do whatever gives you the results you like.
Rich, now that we have in GraXpert background, deconvolution and noise reduction, which of them you do in Siril and which in GraXpert? Waiting for the new SIRIL version to have them integrated, I'm trying to build a kind of standard workflow. Thank you
I've been using GrtaXpert for all of them. Keep in mind that GraXpert deconvolution is non-stellar (object-only) only right now. The next update will include stellar (stars-only) as well.
Dumb Question: After applying background removal, deconvolution and then denoising, when you "save processed" , are all the steps applied to the saved FITS? I ask because i notice the image is named differently when saved when you select the views at the top of Graxpert.
Not a dumb question at all. The 'Save Selected' button will save whichever view you have selected from the top pulldown. Each selection will include the function that you ran previously. So if you ran (in order) BGE, then deconvolution, then denoising, selecting the Deconvolved Object-only image would also include the Background, and selecting the Denoised image would include the Deconvoled Object-only and Background processes.
Brand new to this so... If I use the AI deconvolution in GraXpert, do I still need to use the Starnet Star Removal function in Siril before I start stretching an image?
Thank you very much for the presentation and I am already looking forward to the new version of Siril But one question, the PSF evaluation in Siril shows the average of the selected stars in Moffat or Gauss which makes a difference. In the console, however, a value of FWHN channel 1 is shown at the same time. Do you have any idea what this value represents?
My FWHM is between 13 and 20. I am shooting Orion and deselected orion itself as an object to get down to those numberss. Bigger stars with glow around them are still selected. Is that acceptable number range?
This question is bothering me too. A lot of People on Internet are sugesting to deconv before denoise, because o nature of PSF. But first data must be color calibrated, and i've noted that whis model works only on stretched data. So BGE -> color calibration ->statistical stretch -> deconv -> denoise Thats what i've got, but dont take it as a granted
@@MarekDrozdowski-fi3bh This model works on LINEAR data. A suggested by the ordering in the left menu of graxpert background extraction -> deconv -> denoise
As a MacOS user I hope Siril fixes the annoying cosmetic bug where you choose a tool and only the title bar appears. It doesn't happen every time. Also when bring up Starnet window it sometime seems to take a while for the tool window to display. You have to click on it to display the tool. I hope they can add some performance enhancements for Apple Silicon processors. jI don't know if the GraXpert background is as good as the Siril method. Sometimes I think the file that GraXpert creates after BGE is harder to process. It could be because I'm new at this but I can't seem to get the same result using GraXpert compared to Siril BGE. Hard to explain.
Is that a known issue? I don't run on a Mac so I'm not familiar. Most of the code in Siril for BGE is GraXpert, minus the AI obviously. Not sure what could be happening.
@ It was sharpened the for sure My question if is all about sharpening Can’t we do the same thing on Ps Or on graxpert prevent the noise as sharpening?
hello, I'm new at using GraExpert with WIN 11. I try to denoise, but fails, so I went to the GraExpert Log file,: "The GPU will not respond to more commands, most likely because of an invalid command passed by the calling application." Any clue how to fix this ? Thanks for your insight! I appreciate your time and love your videos
I've not been able to get it to work; first time it rebooted my computer, the subsequent attempts just resulted in error messages. I try a complete fresh install and see if I can get it to work.
@@DeepSpaceAstro Thanks Rich, I updated my graphics driver, complete uninstall(including the models and preferences) and reinstall; that seems to have done the trick.
I thought the video was about showing the de-convolution f the object , yet you zoomed in on stars after and of what I could see of the nebula detail I couldn't see any difference ?
I was trying to show what you could see of the Flaming Star. It's hard to see in the video, and you'll have to play with the settings, but it does work.
@Fractalite I can vouche for its capabilities. I've tried it out, and you can definitely see a difference in detail of the nebula. Give it a go yourself... play with the settings as Rich suggests.
I'm right there with you. I literally could see zero difference between the deconvolved version and the original. I can tell from the rest of the comments, the problem must be on my end. Everyone seems very excited about this. I guess I don't understand what I should be looking for. What does this process actually do?
@@charliehubbard4073 I have used deconvolution in other apps and could never see any difference. I have also watched numerous videos about the subject and it's application and I have never gained any understanding of it or what it is supposed to do whatsoever. I am not trying to discount the feature and acknowledge it may be my ignorance, but I do not get it at this point. I have had no problem grasping the understanding of any other processing functions I have utilized, but every time I have added deconvolution to my workflow I always felt like I wasted several minutes clicking around because I have never been able to notice any improvement to the image.
I think so. I haven't gotten that far yet. I'm kind of expecting another model for that piece in the future. I haven't heard that, just my thoughts since they have a models list with only one model right now.
@@DeepSpaceAstro Sorry! I have just realized. it is FWHM and not FHWM. The measuring util is correct, you need to input the FWHM in pixels, not arcsecs
That may be the memory on your GPU not the system RAM. Try lowering the batch size in the Advanced settings, and/or disabling the hardware acceleration.
@@DeepSpaceAstro thanks. it suddenly and spontaneously started not crashing after I did nothing. Guess one can color me not very impressed so far. I've used everything from maxent to lucy-richardson over the years and this is really the least impressive I've seen so far.
I missed that the average star FWHM is show in arc seconds, not pixels which GraXpert requires. When you click the Average PSF button, look in the console and you'll see a line that looks like this:
Found 15012 Gaussian profile stars in image, channel #1 (FWHM 4.420661)
That FWHM is in pixels, and is what you should use in GraXpert. Sorry about that...
There's nothing to apologize for and it should be transferred automatically in the next version 0I0
In addition, this value that appears in the console was not documented anywhere and was probably incorrectly labeled FWHN, which shouldn't be a problem.
Deleting the focal length doesn't really make sense to me because it only changes the value for FWHMx/y
Super support😁👍
It looks a little different in Siril 1.2.5 when I click average PSF: (What value do I use besides the middle of FWHMx/y?) Here's what is showing:
Average Gaussian PSF
N: 91 stars (3 saturated and excluded)
B: 626.016567
A: 5586.769714
FWHMx: 6.58"
FWHMy: 5.95"
r: 0.905
Angle: 0.14 deg
rmse: 1.191e+02
Sorry, what looks different? Per my comment above you need to use the value show in the console not in that Average Star Data. Or am I misunderstanding?
@@DeepSpaceAstro My bad, it took me a couple reads to understand to look in the console in Siril! Appreciate your quick responses!!!!
@@Waialua-Wingfoil You can also read it in the deconvolution module after the PSF determination.
Just tried it out.... works very well on bringing out nebula details. Then a little denoising ... big improvement. Thanks for the update, Rich. 😊
Glad you're seeing results! Thanks!
Thanks Rich. Great times indeed for our hobby.
Wow! The pace of development of these tools is breathtaking, thanks again for the heads up.
Really cool. Siril 1.4, new features, and GraXpert with new features, all integrated. The future is bright ✨. Very informative, thanks very much from (cloudy 😬) Germany 👍🔭
Couldn't agree more! Thanks!
I wish u clear skies buddy❤
THiS IS WHAT IVE BEEN WAITING FOR
It will deconvolute your caps
Lets play nice now.
@@DeepSpaceAstro was just meant as a lame joke, my apologies to astroduck
Would be great if you can do a comparison between GraXpert and SETI Astro.
Thanks a lot for all your videos
It's my pleasure
Ahh , does make a difference. I didn't notice at first, but I'd forgotten to check my fwhm average in siril. Once I input that correctly into graxpert, there was a marked difference, even with dwarf data
somehow i missed your video but finally caught up. things are just getting really good now.
Having such powerful tools available for free is just nuts - can’t wait to try this out!
Like always excellent friendly video
Thanks so much for your help and update
I haven't, sorry. Thanks!
This is brilliant Rich I'll probably wait for the non beta release before I try it out but I'll definitely be keeping my eye on it
Goodbye pixinsight 😊
nah blurX is peak and pix is overall a better software
@@JLCubing the main downside of pix is that it is designed to look like software from the 90s, its incredibly slow and tedious for simple things (especially layer based editing/why do we have to have a separate preview window for EVERYTHING?)
it can be faster if you learn it well yes but its still designed badly
@@TheOfficialProtostar that's just pixinsight's design language. and the part when you mentioned slow; you need a good pc for it and allat for it to run fast. I dont mind with the extra windows cuz I don't do real heavy processing lol
@JLCubing no when i said slow i meant its slow to operate because of its design not because "you need a good pc" haha, i have a 13th gen i9 and a 4080 my pc isnt slow, and im familiar with many complex softwares
@JLCubing and the extra windows i mean, why do we need to have a "preview window" to show us what a simple damn curve is doing? 99% of photo/visual software overlays it ON your photo and has a way to disable the preview. Even with ways to do layer based merging in pixi its so backwards, why do i need to install a llugin or script to be able to easily use blending modes that also take several seconds to update instead of instantly lmao
Wow! Great feature! One question: with Deconvolution now available in GraXpert, what would be the "proper" workflow? Should I run Crop, Background Extraction, and Deconvolution in GraXpert, then move to Siril for PCC, StarNet removal, and GHS to adjust the background image with proper saturation? After that, should I return to GraXpert for denoising (linear data) and finish with star recomposition in Siril (or Photoshop)? What do you think of this workflow? Before GraXpert's Deconvolution feature, I used to run Siril's Deconvolution right after PCC.
I use GraXpert to run BGE, Decon, and then Denoise. Then jump in Siril for PCC, star removal, stretching, and then add back the stars. That's just my process though. FYI, all 3 functions in GraXpert should be ran while the data is still linear, not stretched. You can do it after stretching if you want though. It's just the recommendation. In the end, do whatever gives you the results you like.
Rich, now that we have in GraXpert background, deconvolution and noise reduction, which of them you do in Siril and which in GraXpert? Waiting for the new SIRIL version to have them integrated, I'm trying to build a kind of standard workflow. Thank you
I've been using GrtaXpert for all of them. Keep in mind that GraXpert deconvolution is non-stellar (object-only) only right now. The next update will include stellar (stars-only) as well.
Very interesting video's Rich, great news from SRIl and Graxpert, hope they will include Seti Astro in the future also
I believe Franklin from Seti Astro will be using the new Python API to bring his apps into Siril when version 1.4 is released.
Dumb Question: After applying background removal, deconvolution and then denoising, when you "save processed" , are all the steps applied to the saved FITS? I ask because i notice the image is named differently when saved when you select the views at the top of Graxpert.
Not a dumb question at all. The 'Save Selected' button will save whichever view you have selected from the top pulldown. Each selection will include the function that you ran previously. So if you ran (in order) BGE, then deconvolution, then denoising, selecting the Deconvolved Object-only image would also include the Background, and selecting the Denoised image would include the Deconvoled Object-only and Background processes.
Brand new to this so... If I use the AI deconvolution in GraXpert, do I still need to use the Starnet Star Removal function in Siril before I start stretching an image?
Yeah they're 2 separate functions. You don't have to remove the stars, it's just what a lot of us do before stretching. It's up to you.
Thank you very much for the presentation and I am already looking forward to the new version of Siril
But one question, the PSF evaluation in Siril shows the average of the selected stars in Moffat or Gauss which makes a difference. In the console, however, a value of FWHN channel 1 is shown at the same time. Do you have any idea what this value represents?
I just pinned a comment explaining that.
@DeepSpaceAstro no problem and thanks for the clarification because the med values that came out of it could not be entered into the field☺
I can´t see any diference 😓
My FWHM is between 13 and 20. I am shooting Orion and deselected orion itself as an object to get down to those numberss. Bigger stars with glow around them are still selected. Is that acceptable number range?
See my pinned comment
@@DeepSpaceAstro Thanks!
Thanks Rich! Very useful review!
Did you have a chance to compare this new feature with deconvolution from Cosmic Clarity?
Thanks! No haven't compared the 2 yet.
Great video - thanks!
Glad you like it! Thanks!
Do you denoise before deconvolute and do both before stretching?
This question is bothering me too. A lot of People on Internet are sugesting to deconv before denoise, because o nature of PSF. But first data must be color calibrated, and i've noted that whis model works only on stretched data.
So BGE -> color calibration ->statistical stretch -> deconv -> denoise
Thats what i've got, but dont take it as a granted
@@MarekDrozdowski-fi3bh This model works on LINEAR data. A suggested by the ordering in the left menu of graxpert background extraction -> deconv -> denoise
Decon then denoise, on linear data.
As a MacOS user I hope Siril fixes the annoying cosmetic bug where you choose a tool and only the title bar appears. It doesn't happen every time. Also when bring up Starnet window it sometime seems to take a while for the tool window to display. You have to click on it to display the tool. I hope they can add some performance enhancements for Apple Silicon processors.
jI don't know if the GraXpert background is as good as the Siril method. Sometimes I think the file that GraXpert creates after BGE is harder to process. It could be because I'm new at this but I can't seem to get the same result using GraXpert compared to Siril BGE. Hard to explain.
Is that a known issue? I don't run on a Mac so I'm not familiar. Most of the code in Siril for BGE is GraXpert, minus the AI obviously. Not sure what could be happening.
Would you still run a second deconvolution back in siril for the stars ,or is that overkill
I would say yes, if that's what you want. Give it a shot. It won't hurt anything.
Hi again
Question I have noticed that it says deconvolution object only
Does it mean stars are not deconvoluted ?
That's correct. I believe they're working on another AI model for deconvoluting the stars.
@
It was sharpened the for sure
My question if is all about sharpening
Can’t we do the same thing on Ps
Or on graxpert prevent the noise as sharpening?
hello, I'm new at using GraExpert with WIN 11. I try to denoise, but fails, so I went to the GraExpert Log file,: "The GPU will not respond to more commands, most likely because of an invalid command passed by the calling application."
Any clue how to fix this ?
Thanks for your insight! I appreciate your time and love your videos
Maybe try disabling Hardware Acceleration. Right-hand side, click Advanced, and the switch is at the bottom.
@@DeepSpaceAstro Bingo! It worked!
I've not been able to get it to work; first time it rebooted my computer, the subsequent attempts just resulted in error messages. I try a complete fresh install and see if I can get it to work.
If you can't get it going, they have support on Discord.
@@DeepSpaceAstro Thanks Rich, I updated my graphics driver, complete uninstall(including the models and preferences) and reinstall; that seems to have done the trick.
Hi Rich! Is this something similar to what BlurX does?
I think it's similar, but not sure if it actually works the same.
@@DeepSpaceAstro With M16 work fine!
I thought the video was about showing the de-convolution f the object , yet you zoomed in on stars after and of what I could see of the nebula detail I couldn't see any difference ?
I was trying to show what you could see of the Flaming Star. It's hard to see in the video, and you'll have to play with the settings, but it does work.
@Fractalite I can vouche for its capabilities. I've tried it out, and you can definitely see a difference in detail of the nebula. Give it a go yourself... play with the settings as Rich suggests.
I'm right there with you. I literally could see zero difference between the deconvolved version and the original. I can tell from the rest of the comments, the problem must be on my end. Everyone seems very excited about this. I guess I don't understand what I should be looking for. What does this process actually do?
@@charliehubbard4073 I have used deconvolution in other apps and could never see any difference. I have also watched numerous videos about the subject and it's application and I have never gained any understanding of it or what it is supposed to do whatsoever. I am not trying to discount the feature and acknowledge it may be my ignorance, but I do not get it at this point. I have had no problem grasping the understanding of any other processing functions I have utilized, but every time I have added deconvolution to my workflow I always felt like I wasted several minutes clicking around because I have never been able to notice any improvement to the image.
@@astroshooter1960Ok . I'll have to wait for Siril 1.4 to come out .
Love these videos mate, they’re so helpful. Thanks so much 😊
Glad you like them! Thank you!
Does it improve stars like Bxt does?
They said this AI model leaves the stars unaltered.
@@DeepSpaceAstro Then we would need to use Siril's deconvolution for the stars after using the GraxPert DC for the nebula?
I think so. I haven't gotten that far yet. I'm kind of expecting another model for that piece in the future. I haven't heard that, just my thoughts since they have a models list with only one model right now.
@@DeepSpaceAstro Yes, there will be another model for stellar deconvolution that corrects aberrations too :)
Wait but siril is showing FWHM in arcseconds, not pixels
You're right! I wonder if that's a typo in GraXpert? Not sure. I also noticed the spelled FWHM wrong.
I saw that missspelling too :D
I made a test on starnet RGB sample, because my own data sux, and it looks good at 6.5px and 1.0 power. So I think they really mean fwhm in pixels
@MarekDrozdowski-fi3bh noticed that, too... I think it's just a typo.
@@DeepSpaceAstro Sorry! I have just realized. it is FWHM and not FHWM. The measuring util is correct, you need to input the FWHM in pixels, not arcsecs
Nice!
Yes!!
PixInsight is going to have to dramatically reduce it's overpriced software to stay in business.....
Yee Haw! 😁
😂
humans are rad.
Crashed EVERY single time I tried it. error in log was out of ram. seems unlikely as I have over 7 gbytes available. any suggestions anyone?
That may be the memory on your GPU not the system RAM. Try lowering the batch size in the Advanced settings, and/or disabling the hardware acceleration.
@@DeepSpaceAstro thanks. it suddenly and spontaneously started not crashing after I did nothing. Guess one can color me not very impressed so far. I've used everything from maxent to lucy-richardson over the years and this is really the least impressive I've seen so far.
I don't see a difference. What's going on?
You should. It can be subtle depending on your data and settings used.
you go too fast for me... you sound like the voice that reads the list of serious issues at the end of some TV ad for some drug ...