Overview
- ESA researchers David O’Ryan and Pablo Gómez built the neural network AnomalyMatch to detect rare objects in decades of Hubble data.
- Scanning nearly 100 million image cutouts in about two and a half days, the tool produced a ranked list that human review confirmed as more than 1,300 true anomalies.
- Finds include numerous galaxy mergers and interactions, gravitational lenses, jellyfish and ring or clumpy galaxies, edge-on planet-forming disks, and several dozen objects that resist classification.
- This work constitutes the first systematic anomaly search of the Hubble Legacy Archive and is documented in a peer-reviewed Astronomy & Astrophysics paper.
- The team highlights the method’s scalability for upcoming data-rich missions, with detailed classification and scientific follow-up of the most intriguing candidates to come.