AI Tool Flags 800 Undocumented Hubble Objects in 35-Year Archive Search

1 articles · Updated · spacedaily.com · Jun 6

Nearly 100 million Hubble image cutouts were screened in about 2.5 days, producing a shortlist that astronomers narrowed to more than 1,300 visually anomalous cases and a catalog of 1,255 unique objects.
More than 800 of those objects had not been described in scientific literature, but the study says they are candidates rather than confirmed discoveries because researchers still had to verify the AI-ranked images by eye.
The catalog is dominated by more than 400 merging or interacting galaxies and includes 86 new gravitational-lens candidates, plus ring galaxies, jellyfish galaxies, star-forming clumps and some edge-on planet-forming disks.
Several dozen objects did not fit existing classifications at all, making them prime targets for follow-up observations that will test whether the most unusual candidates hold up.
ESA says this is the first systematic anomaly search of the Hubble archive, offering a trial run for AI-assisted screening as Euclid and the Vera C. Rubin Observatory generate far larger image troves.