Astrobin's new restrictions on artistic extression

Thanks for the clarification. The part in parentheses seemed to amplify the restriction in the first part of the sentence. I am satisfied that Astrobin remains a place for me and other art-wise image processors.

Alex

It's interesting to note that Starnet ++ (for star removal) would be an example of AI that a lot of use in processing astroimages. So good thing that's allowed! 

Sigga:
Is not removing all stars from image "made up", no stars not represent reality in any part of sky I ever see.

It's not made up. It's a real picture of a real object, that has been altered. And let's not get philosophical with the ship of Theseus, it's really quite basic common sense

I see the debate, but I am more on the AI/artistic side of the discussion.  Clearly you are saying that AI is OK in Starnet or Topaz to process real images… heck half of PixInsight could be considered AI…

That being said I do support Alex's perspective since art for art's sake will set standards of how we process regular images in the future… why not just a separate category like Equipment is a category….

It's a hard topic. In fact, I think all our works from this hobby are somewhere in the middle of "astrophotography" and "abstract space art", some are closer to point A and some are closer to point B. The post-processing we are doing is actually abstracting and altering the data to make the subject more pronounced and prettier, isn't it?

"Common sense" is the keyword here.

No, no, it's not the sin of AI tools. Only with PixInsight or any "traditional" processing tools, the stuff could be made out of nowhere - The same thing as "It's not the camera, it's the mind behind it."

Alex,

Clearly you misunderstood Salvatore, if anything he gave me the impression during the debate that he is too liberal, not at all too restrictive. No one during the debate talked about banning art, and most of your images are clearly artistic. 

The debate was not about banning, but about ethics and transparency :  revealing that you used AI while processing your image. The traditional softwares we use are not AI. Conflating everything together as "AI" does not make any sense. Pretty soon people will argue that Registaxx was using AI and Topaz is not an inovation. 

The unfinished debate (too bad it slipped again towards the award Astrobin system) was about people artificially enhancing their image, with unreal and beyond their system details, while hiding the fact that they are the result of AI and presenting the image as any other classical astrophotography. That's gaining an advantage and deluding the viewer (and themselves), while being dishonest about what part comes from their equipement and what part is artificially produced. Especially when no one thinks their images are "artistic" in the same sense yours are. 

Starnet has been pointed out as a community accepted AI and as a precedent for therefore allowing liberal (and hidden?) use of all present or future AI. Starnet is simply a tool for extracting stars, with the advantage of sometimes being more performant and more complete than manual star extracting tools. Sometimes it's worse. Anyway, it does not produce any new details that increase the resolution or make the image look better. If anything it makes the image lose details, introduces artifacts where bright structures are mistaken as stars and looks grainier. No one can be fooled that a starless image (with all those grey spots) is the real product of your equipement. So no, Starnet does not bring an advantage in the same sense Topaz does. 

Your images and authentic artistic productions are pretty safe. We are not safe from better than reality images hidden under the umbrella "hey, it's just art".

All best

Bogdan

Salvatore Iovene:
This Reddit post shows the results of an AI trained on AstroBin images (amongst others) to generate a completely made-up nebula:

https://www.reddit.com/r/Astronomy/comments/mge6lf/this_nebula_does_not_exist_made_with_styleganv2/

More details in the comments on Reddit.

Thanks for sharing that example. I had never seen an AI-created DSO that used actual amateur astrophotography for it's training model. I definitely see the influence of many of the most-popular nebulae we amateur astrophotographers capture in there.