Seti Astro has just released a new stand alone AI based sharpening/deconvolution tool called Cosmic Clarity

[PhilB61]

Just watched a Seti Astro YouTube video regarding the above and downloaded the software.

Currently in Beta form so a bit of a clunky interface and only works on Windows and with tiff files. Good news is that it works totally stand alone, no Pixinsight or Photoshop etc required.

I have already tried in on a couple of images and I think the results are very impressive, I will do more testing but these initial results suggest it works better than any of the other free tools which I have used and also comparable with the results I am getting from my Pixinsight/BlurX trial (which is a huge complement as I think these produce amazing results).

I continue to be impressed and amazed by all the free (and paid for) tools which continue to be developed for our hobby by some very talented people out there, thank you.

 

[Ouroboros]

Interesting thanks. Yes, it’s amazing. These tools are coming thick and fast at the moment. It’s almost difficult to keep up with them all. If past form is anything to go by I expect Seti Astro will make this new tool available for other platforms and Pixinsight etc.

[Stuart1971]

2 hours ago, Ouroboros said:

Interesting thanks. Yes, it’s amazing. These tools are coming thick and fast at the moment. It’s almost difficult to keep up with them all. If past form is anything to go by I expect Seti Astro will make this new tool available for other platforms and Pixinsight etc.

It has a PI process already and is being sorted for MAC users too…👍🏻

[Ouroboros]

2 hours ago, Stuart1971 said:

It has a PI process already and is being sorted for MAC users too…👍🏻

Excellent.

Processing in PI increasingly involves using a succession of readily applied processes and scripts provided by third parties - some paid for and some not.   I am increasingly wondering how this works commercially. Do main third party contributors get paid in some way? It would seem unfair if they don’t benefit in improving what is after all a commercial product. 

[Stuart1971]

14 minutes ago, Ouroboros said:

Excellent.

Processing in PI increasingly involves using a succession of readily applied processes and scripts provided by third parties - some paid for and some not.   I am increasingly wondering how this works commercially. Do main third party contributors get paid in some way? It would seem unfair if they don’t benefit in improving what is after all a commercial product. 

I think it is a bit of a sore point with the PI developers, I know they had big issues with the guys at GraXpert when that came out, and tried everything to stop it working with PI, but came to an understanding in the end, they do work closely with Russ Cromann, not sure about other script developers, but I am pretty sure none of them get any money for there scripts..they all improve PI, BUT god forbid you mention any of them on the PI forum, they will just warn you then ban you, that is an absolute no no, so that says it all really…I fell foul of that and was warned, after asking a question about GraXpert..

[TiffsAndAstro]

Tried this tool a few mins after his video.

Doesnt work with some PNG or 32bit tiff of same image. Also anything above 0.3 for stars results in pacman shapes.

At least for my so so images/data

Overall I think it has potential but I find it hard to know for sure right now

Very glad of his efforts though 

This one looks ok, few odd newvstar like artifacts along edge of my wall?

Edited 15 hours ago by TiffsAndAstro

[Ouroboros]

52 minutes ago, Stuart1971 said:

I think it is a bit of a sore point with the PI developers, I know they had big issues with the guys at GraXpert when that came out, and tried everything to stop it working with PI, but came to an understanding in the end, they do work closely with Russ Cromann, not sure about other script developers, but I am pretty sure none of them get any money for there scripts..they all improve PI, BUT god forbid you mention any of them on the PI forum, they will just warn you then ban you, that is an absolute no no, so that says it all really…I fell foul of that and was warned, after asking a question about GraXpert..

Interesting. Thanks for the warning. 

[Ouroboros]

@TiffsAndAstro I guess you have but have you let Mr Seti astro know? Sorry but he doesn’t appear to have a name. 

[TiffsAndAstro]

4 minutes ago, Ouroboros said:

@TiffsAndAstro I guess you have but have you let Mr Seti astro know? Sorry but he doesn’t appear to have a name. 

its only beta and it seems impressive but with a few foibles. might try and post on his yt video. 

also its very possibly its a problem with my images/data. i still get sn++ via siril causing a blue screen of death (stop_clock) occasionally but only with my data - never anyone else's 

Edited 14 hours ago by TiffsAndAstro

[The Lazy Astronomer]

5 hours ago, Stuart1971 said:

I think it is a bit of a sore point with the PI developers

I think it's just that it's against their ethos,  generally, to use any kind of 'black box' tool, and I kind of get that they don't want their own forum filled with comparisons of their tools vs 3rd party ones, but it is a tad drastic to restrict any conversation at all, even just people seeking help, but whatever, plenty of other forums. 

[symmetal]

4 hours ago, Ouroboros said:

I guess you have but have you let Mr Seti astro know? Sorry but he doesn’t appear to have a name. 

His homepage does say 'photos & videos by franklin marek' and it's also the user name on his Astrobin page. 😊

I've used his Auto DBE script which works very well on most images with just one click without having to set any control points manually.

Alan

[Ouroboros]

37 minutes ago, symmetal said:

His homepage does say 'photos & videos by franklin marek' and it's also the user name on his Astrobin page. 😊

I've used his Auto DBE script which works very well on most images with just one click without having to set any control points manually.

Alan

OK. Thanks for telling me his name. 

Does his Auto DBE script somehow avoid stars and nebulous bits?   I must say that’s something I wonder about these automatic gradient removal tools. I mean …. it’s often difficult to tell by eye what’s gradient and what’s faint nebulosity.  So how a script can distinguish between the two I  don’t know.

Of late I have been deferring DBE until later in the processing - after star removal in fact and even then being very careful where I drop the crosses.  Sometimes I have not even bothered with it unless there’s some very obvious gradient.  

[symmetal]

@Ouroboros Yes it does. If you watch his Youtube video describing how it works, he compares it to GraXpert and points out with some examples that GraXpert can remove some nebulocity which ADBE didn't. On areas with dark 'nebula' he recommends masking them out before running it, which is likely why is didn't work well on some of my images.

Usually it's one-click and 10 secs to run. If it doesn't do what you want just click 'undo' in PI. 🙂 Better than strugging for 30 mins or more moving control point around to get DBE to work.

Alan

Edited 7 hours ago by symmetal

[PhilB61]

Ok, so I had a bit more time to play with this new tool. No problems opening any of my tiff files from various sources, Siril, PI etc although nothing much bigger than 300mb.

Possibly I was a little overly enthusiastic in my initial assessment, my excuse is that it was in the early hours. I still think it is a great free tool but as Frank himself says in the video, its not in the same league as BlurX. Comparing side by side versions of the same image, the BlurX versions (on default setting) are both sharper, cleaner and less noisy than the best obtained using optimised settings with Cosmic Clarity. I also noticed as mentioned in one of the posts above that using a high value for Stellar sharpening introduced a hint of false stars on the boundary of some high contrast areas but only noticeable/visible if really pixel peeping.

Frank does mention that he will be continuing to develop the AI model so I am sure it will be even better once out of beta. 

 

[TiffsAndAstro]

1 minute ago, PhilB61 said:

Ok, so I had a bit more time to play with this new tool. No problems opening any of my tiff files from various sources, Siril, PI etc although nothing much bigger than 300mb.

Possibly I was a little overly enthusiastic in my initial assessment, my excuse is that it was in the early hours. I still think it is a great free tool but as Frank himself says in the video, its not in the same league as BlurX. Comparing side by side versions of the same image, the BlurX versions (on default setting) are both sharper, cleaner and less noisy than the best obtained using optimised settings with Cosmic Clarity. I also noticed as mentioned in one of the posts above that using a high value for Stellar sharpening introduced a hint of false stars on the boundary of some high contrast areas but only noticeable/visible if really pixel peeping.

Frank does mention that he will be continuing to develop the AI model so I am sure it will be even better once out of beta. 

 

I think it still looks very promising for those without pi though. Surely better than having nothing?

Does it overlap or interfere with astro sharp and astro clean?

[PhilB61]

I agree a very promising start, and with the prospects of further improvements to come. Don't know how it compares/complements or otherwise with AstroSharp etc as I've had little success getting it to work at all.

 

[ONIKKINEN]

Gave it a test, cant say i am too impressed at this point. But its a first release so it can only get better.

GIF of 3 different sharpening methods below. The Seti astro tool at nonstellar sharpening of 2, which is way too much and frankly does a bad job (larger levels only get worse). Then the tool again but at a level of 1, which is much better. Last frame is simple sharpening in photoshop, no AI tools or plugins, just the smart sharpen done in a minute at just one level.

To my eyes the simple photoshop sharpening filter wins this comparison. The Seti astro tool also makes some mistakes, i think it doesn't quite understand galaxy detail and has turned some blobs of galaxy "stuff" into perfect round stars. The tool has also attacked the background and made it blotchy, which is common for sharpening tools but not what you want to see in a tool like this. It does reduce star sizes as much as you want to, in this experiment i ran it at 0.4. But i have to be honest here, star sizes can be made whatever we want them to with simple starXterminator or Starnet++ workflow, by simply not stretching the stars too big in the first place.

[Ags]

I quite like what it does.

Before:

50% sharpened (stars and nebula):

90% sharpened: