BlurXTerminator: AI Version 4

Russ Croman has updated BlurXTerminator with AI version 4. In this video Russ and I discuss and demonstrate its new performance capabilities. Needless to say... it is pretty amazing.
The Release Notes can be found here:
www.rc-astro.com/blurxtermina...
Purchase BlurXTerminator and more Now:
BlurXTerminator
www.rc-astro.com/software/bxt...
StarXTerminator
www.rc-astro.com/software/sxt...
NoiseXTerminator
www.rc-astro.com/software/nxt...
Chapters:
00:00 Introduction
03:00 Russ Explains
04:27 Release Notes
04:55 Aberration Correction!
07:17 Linear Images
08:30 Flux Conservation
09:45 Example 1: Widefield Aberration Correction
16:34: Correct First then SPCC
18:00 Example 2: Detecting Pease 1 in M15!
22:11 Stellar and Non-Stellar distinction
24:23 The Reveal
26:54 Example 3: M88 Through the Versions
32:02 Example 4: NGC 2392
More at:
AdamBlockStudios.com
See the Reviews:
www.adamblockstudios.com/cate...
Those just starting out in PixInsight will enjoy FastTrack Training!
(Updated to include SPCC, FastIntegration and more!!!)
www.adamblockstudios.com/cate...
(see the video • Introducing FastTrack ... )
#PixInsight #BlurXTerminator #NoiseXTerminator #StarXTerminator #Deconvolution #neuralnetworks #M15 #NGC2392 #NGC1491 #AdamBlockStudios

[@JoeBob79569]

It's crazy being able to see a planetary nebula like that from 40,000 light years away. At these distances my nerdy brain always go back to Star Trek Voyager, which ended up 70,000 light years from Earth; and being able to see something this far away, from your own back yard, is kind of mind boggling.

[@AdamBlock]

RIght?... thanks for watching

[@TomPeter-xn2ib]

Thank you SO MUCH Russ for the release of AIV4, and thank you Adam for your comprehensive demonstration of it's use! I use an NP127is / ASI6200MM and ever since I got the camera, I've been fighting with getting the stars in the corners correct....it's close but not perfect after 2 1/2 years of adjusting. The first time I used the new AI4.....PERFECTION!!!! No more fiddling with minute adjustments!! Russ you are a genius as are you Adam!! Thanks so much to both of you!!

[@DamonScotting]

Well done, Adam! Once again you have done an amazing job of breaking it down. I'm currently going through some of my old shots and testing AI 4 out, very impressed so far. As always, Russ, you're an Astro wizard!! 👏🧙‍♂

[@AdamBlock]

Unfortunately I am not doing a good enough job based on the real world response. There are others saying half as much (and explaining far less) with equivalent or far greater results. Kinda of a harsh world in this KZbin verse as you are "graded" with each video you make.

[@gregerianne3880]

Thanks for showing us these new capabilities, Adam, and a thousand thanks to Russ for all his hard work and making so much more possible for us! Excellent.

[@vivekbannore2250]

BXT 2.0 is awesome !!! I wish I could post a before and after using the initial Correct Only mode. It literally made me say "Shut up and take my money" but then I realised I already own BXT and not running a trial version 😁. People its a game changer. AI v4's optical correction is out of this world. Thank you Adam for your video and taking time to explain everything. Love your videos and deep dives. Russ thank you so much for your hard work and bringing us this wonderful and out of this world update. RC-Astro rocks !!

[@antoniozaurrini1694]

Thank you so much Russ and Adam for taking the time to show us the before and after and just in time for Christmas, i also cant wait to update BXterminator, keep up the work guys ! :)

[@aleksandartodorovic6563]

Thanks Adam & Russel, for the presentation, can't wait to give this a try! Happy Holidays!

[@davidswinnard7565]

Brilliant! Thanks Russ for these tools, thanks Adam for the in-depth explanations.

[@frl8031]

Thank you both so very much! You're just wonderful. We live in the best age in the history of humanity!

[@astrodug]

Thank you Russ and Adam for this video and for the tools to improve our images. As soon as I watched this video, I updated BlurXterminator and gave it a try. All I can say is WOW!!! I tested it on some data from my old 10in f/6.3 LX200 and an ASI2600MC Pro. The f/6.3 LX200 is known for coma, and in this particular image, it was out of collimation a bit. Two mouse clicks later and all the deformed stars at the outer corners were corrected. Russ, you are a genius.

[@franksteffens6502]

Thank You Adam and Russ, I just updated to AI 4 and did some tests on old data, what an Incredible performance boost!

[@WestwoodAstro]

I just gave AI 4 a spin and it's absolute, flippin' magic! Thank you Russ for the update and Adam for the terrific overview!

[@josephluciani5531]

Brilliant review! Thank you Adam for this, much appreciated.

[@terrylovejoy3147]

This is fabulous work from Russ. Can't wait to get this and try it on some old files.

[@Si-fp2ij]

Wow Adam, Pease-1 blew me away congrats to Russ on the new update of blur xt Can’t wait to try it!!! Cheers Simon

[@TheMje1963]

There's a lot of haters out there who rag on Russ's software, I'm a lover of his stuff and how much time is saves me, especially since i'm one of those who could never get decon to work right for me without the worms. I have been going over some of my data with the latest version and am loving the results.

[@duendepiola6742]

If you like their software, there are also image generators with AI, so you avoid buying a telescope, mount and accessories.

[@TheMje1963]

@@duendepiola6742Case in point, there are always haters out there just like you!

[@solati78]

@@duendepiola6742great tip, I’ll look into it, thanks!!

[@davewilton6021]

Thank you Adam and especially Russ. I've got a ton of data taken with an out-of-collimation refractor (now fixed) that I'm now eager to reprocess.

[@darkmatterastro]

Thanks Russ and Adam. Because of Russ' tools I'm taking data from three years ago reprocessing them and being amazed at the results. Looking forward to what you have planned for AtarXterminator.

[@davidecattani8282]

Thanks so much Adam for the unparalleled scientific approach you put into your explanations, so much appreciated! And thanks to Russ as well for his unvaluable work, able to extract more and more wonder out of our own images!

[@AdamBlock]

Thanks!

[@jonathanpearceff]

Thank you Adam and double thank you Russ!

[@giampieroiannone9202]

Thank you from me, too! It is truly magic, even bad coma gets corrected! And thank you for lowering the price on some of the apps, I will be sure to buy more!

[@DSOImager]

Great interview and demo Adam. I think the key word in the whole video was "fun" :)

[@schandler52]

Maestro ?? it’s not Bernstein it’s Russell!! Once again, a fantastic update, and a great pre-holiday present. thanks Russell. Happy holidays.!!

[@TyroTim2000]

Thank you. Well done!

[@AdamBlock]

Thanks for watching!

[@afryhover]

Thanks for this, well done RC and Adam 👍👍👍👍

[@AdamBlock]

Thanks!

[@petermonson3719]

Thank you. Technically helpful but also really enjoyable.

[@AdamBlock]

Excellent..thanks!

[@pksacoma]

Wow congratulations I’ll try it obviously great project 🎉

[@CrispyImages]

I love this! RC is a god among men. Can’t wait to run the new AI on all my terribly aberrated masters. Keep it up 👍🏻

[@GidgyGirl]

Scope manufacturers need to bundle BXT AI4 with their less than or equal to 46mm max image circle scopes for use with modern Astro cameras!

[@AstroBackyard]

I can't wait to try this out!

[@AdamBlock]

Thanks for watching. I still think we should do a project together.

[@AstroQuest1]

Fantastic, thanks Russ for the new version and thanks Adam for the insightful PI usage - could not come at better time as I am activating my Edge SCT for galaxy season and the stars with that scope just aren't at the same level as with my refractors - in my opinion. Cheers

[@richardshagam8608]

Oh, great!! Now I've got to reprocess all of my old images. (I did try it on an already 'completed' and stretched image and it improved that like I'm not supposed to)

[@tomatomaeg]

just played around with some data i had with really bad corners. Updated blurx, applied to a raw stack and my jaw dropped to the floor. THIS IS BLACK MAGIC

[@AdamBlock]

Agreed

[@tomatomaeg]

looks like i don't need to image separate f/4 stars anymore with my canon 200mm f2.8. This is the greatest of the great xmas preset for us telephoto lens widefield guys.@@AdamBlock

[@AdamBlock]

@@tomatomaeg Indeed... this really helps all of those wacky aberrations!

[@RyanParle81]

Thanks Russ for almost completely demotivating me to fix my sensor tilt 😂

[@trevorgreen2232]

Thanks Adam and Russ for a great presentation . It was said to do a correct only of stars first but I wondered if you should do a back ground extraction first or do that after the correction then prior to SPCC ?

[@SpaceTonight]

As an astronomer I have been popularizing astronomy for years, many young people who have no talent for science and are interested in art, photography and astrophotography regret that these photo editing programs are so difficult to use that it discourages them from further developing their passion. If we want to encourage new people to admire and photograph the Universe, we need to take a different direction, easier for them. Dear Adam, you have fantastic knowledge and thank you for sharing it, because I use it myself, but I hope that someone will one day create a program that will process our astrophotos in a few clicks in a satisfactory and friendly way for everyone.

[@haiderbhogadia4829]

Absolutely fabulous Russ and Adam for showing how it works. Cannot wait to update BXterminator. Bw Haider

[@Astro_Shed]

Absolutely superb video, and that to you both….👏🏻👏🏻 Any chance of a Gradient Xterminator in AI to better the new GraXpert ..? I have the old PI plugin which is superb…

[@fazergazer]

Adam, thanks for this interview. I wanted to ask if you know if the PSF is still the “average” PSF or is it spatially variant? Of course the PSF actually varies throughout the image. Thanks!❤

[@redabdab]

13:48 Adam, you used to run PSF Image to get the FWHM value and then plug that value into BXT. But now I see you are checking the 'automatic PSF' box. Can I ask why you made that change? It would sure be easier to do it this way as PSF Image is soooooooo slow! Thanks

[@AdamBlock]

This isn't a change. You can use the manual PSF when there are not enough stars in the image. BXT tiles an image and measures stars within each tile. When tiles are starless you can get slightly different answers- so manual PSF is a uniform solution applied everywhere. It is an exception that there are not enough stars in an image. Starless images of course require manual as well. Otherwise, automatic is definitely the default.

[@redabdab]

@@AdamBlockok. Got it. Thanks for clarifying Adam

[@garyseven4370]

Another great and exciting advance. The only downside is I will now have to re-process all of my old work ;-) One important question comparing the 2022 interview and workflow to the 2023 version: Last year, it seemed to me the Cardinal Rule was "no double-decon". And that was not a limitation of BXT, but of the Deconvolution process itself. Now, it seems to be the recommended workflow: BXT Correct Only -> SPCC -> BXT Again. Am I understanding this correctly?

[@AdamBlock]

Yes. The decon are working in different ways. One gets you less aberrated PSFs... when you can then do further sharpening (though likely you will not need to be as aggressive as previously). BXT correct only isn't *always* necessary. But with data that has lots of aberrations (widefield tends to)... it can be helpful.

[@gordonbulger4965]

I see you use a image comparison tool with a vertical slider. Where do I find this app? Thanks Gordon

[@Stu4man]

Adam, at what point do you do background extraction? Do you do it before the first "correct only" run of BXT or after SPCC?

[@AdamBlock]

That is another can or worms...the winds are changing to do it after SPCC. However, doing it immediately before or after for additive gradients doesn't matter too much.

[@nikivan]

Do you think it would be beneficial to run BlurXT on the individual subs, before the integration?

[@AdamBlock]

Not in general. BXT relies on good S/N...and this is what integrated images have. Integrating images with BXT applied would not give you the same quality result.

[@persianartist]

So if we already have the blurr exterminator would we have to purchase this new version again?

[@AdamBlock]

No. This is an update for existing customers.

[@dgodiex]

no

[@msacco]

Great video but it would've been nice to get Russ to talk more about the new stars correction. Fixing CA is one thing, but this new tool is pretty crazy, too crazy even. It corrects out of focus stars, incorrect backfocus, start trails, etc... It's amazing, but how does it work exactly? Is it just completely re-creating the stars? Does it use your own data to fix it? I think AI is going great, but for the first time it feels like a completely black box that I have no idea what it does and is doing magic. With the main data I know what it's doing - deconvolution, but what happens here is something that I would love to understand, because I wonder if it crosses the line between utilizing AI to fix small things and have AI just invent things. It would be nice to have this addressed somehow.

[@rossc383]

Of course, it generates. Otherwise, it is not possible

[@MisterHyster]

@@rossc383 The question is how and based on what.

[@AdamBlock]

You guys (@rossc383 , @MisterHyster ) need to watch the first video on BXT and read the main release notes. If it were inventing things (in the way I think you mean it) I could not have demonstrated what I did. @msacco Russ does have a couple of technical talks on-line concerning how the neural net is trained and does its thing.

[@rossc383]

@@AdamBlock can you reproduce the same star correction manually as shown in the video? Would love to see the video on this

[@msacco]

@@AdamBlock I have already seen the first video, and that's exactly my point because it's really 2 different things. With deconvolution I know what it does sort of as he explained it, and it is a mathematical process somewhat. With the star correction I tried it crazy scenarios and it completely fixes star trails like here for example: i . imgur . com / Y8nICro . png (remove spaces) He mostly talked about the deconvolution part, but regarding the stars he didn't really explained much which I think would be nice to know and understand how it does it and how accurate it attempts to be.

[@dolphinpat1]

I Have Blurxterminator but not this update. Am I supposed to do something to get it?

[@AdamBlock]

In PixInsight you should check for Updates (manage repositories). Then you need to select the AI version 4.

[@dolphinpat1]

Thanks. Had to download the latest PI version first (don’t know why I didn’t get it). Then discovered I had lost all of my RCAstri processes. Reloaded them one by one (plus Starnet). Took a while but now everything is back to normal. And I have the A4 BlurxT.

[@ivanrovato3588]

*** Error: ERROR: MLLoad() could not load tensorflow graph. Error 3: Converting GraphDef to Graph has failed. The binary trying to import the GraphDef was built when GraphDef version was 440. The GraphDef was produced by a binary built when GraphDef version was 1645. The difference between these versions is larger than TensorFlow's forward compatibility guarantee. The following error might be due to the binary trying to import the GraphDef being too old: NodeDef mentions attr 'explicit_paddings' not in Op output:T; attr=T:type,default=DT_FLOAT,allowed=[DT_HALF, DT_BFLOAT16, DT_FLOAT, DT_DOUBLE, DT_INT32, DT_INT64, DT_UINT8, DT_INT16, DT_INT8, DT_UINT16, DT_QINT8]; attr=ksize:list(int),min=4; attr=strides:list(int),min=4; attr=padding:string,allowed=["SAME", "VALID"]; attr=data_format:string,default="NHWC",allowed=["NHWC", "NCHW", "NCHW_VECT_C"]>; NodeDef: {{node BXT/max_pooling2d_14/MaxPool}}. (Check whether your GraphDef-interpreting binary is up to date with your GraphDef-generating binary.). IS GIVING ME THIS ERROR!

[@CyrilNoger]

I experience the exact same error after running the AI4 :(

[@ivanrovato3588]

@@CyrilNoger just solved it! Write to the support, they'll give you a file zip and you just have to replace a file .dll

[@CyrilNoger]

@@ivanrovato3588 , I've noticed that the BlurXTerminator.4.pb file is half the size of the previous version ! I'll send an email. Thanks

[@stevewilson1935]

'''Me too

[@AdamBlock]

@@CyrilNoger Yes, it is smaller due to some of the improvements (especially the linear processing I believe).

[@yangotira]

AI v2 reduces star sizes better than AI v4

[@Dexterblows]

Adam Block interviewing himself and referring to himself in 3rd person. Adam Block special things? Really?

[@haiderbhogadia4829]

Absolutely fabulous Russ and Adam for showing how it works. Cannot wait to update BXterminator. Bw Haider

[@AdamBlock]

Thanks for watching!