Guest post from Zooniverse participant John Phillip (@john.phillip):
About 15 years ago, I got involved with applications of Artificial Intelligence (AI) in predictive maintenance, as a Mechanical Engineer and Programmer. Since then, I have kept a close eye on AI development. Several years ago, I joined GitHub and started running and training AI models at home to process images and for other random applications.
As I’ve also been an amateur astronomer since childhood, which eventually brought me to Zooniverse almost 15 years ago, it didn’t take long to try AI models for post-processing astronomy images. However, my first try on JWST deep field completely blew my mind when compared to what I could already extract from Hubble images, other observatories and non-astronomical pictures, which prompted me to ask researchers at Zooniverse whether it’s of any scientific value at all. AI enhancement has shown to me that even aging telescopes, like Hubble, can get a boost.
The picture I’ve selected for this Daily Zooniverse post shows a photogenic spiral galaxy that only occupies a tiny fraction of the field of view, but by using AI models trained on noise reduction and super resolution, and regular photo viewing software, it was possible to clean up most of the noise and then magnify the image over 9 times, when compared to the original JWST deep field TIFF picture.
The final result is stored in a Google Drive folder (https://drive.google.com/drive/folders/1e4nSW1-4C5x8jDpJGHDa_9fXlPEzZifb?usp=sharing), together with a few other objects I have processed from the same JWST deep field picture for discussion. A general explanation of the folder contents can be found in the forum (https://www.zooniverse.org/talk/13/2528062?comment=4138256).
Have fun,
John Phillip
Daily Zooniverse 