The Digital Press

Researchers have developed a deep learning pipeline to identify strong gravitational lenses in the UNIONS survey, discovering over 1,300 new candidates. This represents the first multi-band search for such lenses in this survey, combining machine learning with human validation and spectroscopic follow-up.

Why it matters: Strong gravitational lenses are critical tools for studying dark matter, galaxy evolution, and cosmology.

The big picture: Using a ResNet model on 8 million galaxies enables large-scale, efficient multi-band lens discovery in ground-based surveys.

Stunning stat: The team identified 1,346 new strong lens candidates, including 146 high-confidence (grade A) systems.

Quick takeaway: Spectroscopic data confirm 283 lens redshifts and seven high-redshift (z > 2.1) lensed sources, enhancing candidate reliability.