My last post was an #astrophotography image where I framed the Heart and Soul #nebula together with the Double-Cluster. The purpose of that shot was primarily to collect new data to correct the star field for a poor imaging session I had previously where the #Comet C/2017 T2 PanSTARRS was passing through this frame. It was windy and very cold & I missed focus and had ugly stars.

Now thanks to modern #AI tools that can correct for these aberrations, this old data stands on its own all by itself.

For the curious, here is the exact same raw data without the AI based star correction and deconvolution. The algorithms are trained exclusively on Astronomy images to understand and correct ideal PSFs and not “invent” detail that isn’t present in the data itself. You can really see here in this comparison how much of a difference tight and controlled stars make in a good astro-photograph
@mintakka What modern AI tool have you used??
@mintakka always curious how this AI was trained. How was the dataset curiated? Hope the developer publishes samples of the dataset. Just be sure there is no funny business going on.
@HaveyDeckAstro agreed! The Developer is an astrophotographer himself, so im inclined to give the benefit of the doubt. Also in my own testing it seems to live up to his claims that it’s not inventing detail - at least not any sense beyond traditional deconvolution algorithms. Still, I agree it would be nice if he released a little more info regarding exactly how his model is trained.