The Neo paper detailing its application to Rubin data has been accepted for publication in the Astrophysical Journal, underscoring its peer-reviewed status and potential impact.

In testing, Neo improved measurements of galaxy shapes and structures by factors of two to ten compared with standard techniques, potentially allowing Rubin to extract more scientific value without extra hardware.

Neo is a conditional GAN with two networks: one to generate improved images and another to assess their quality.

Rubin Observatory, designed to survey the sky every three nights, benefits from enhanced image quality through AI to maximize scientific returns on its large-telescope investment.

Neo aims to bring Rubin Observatory images closer to space-telescope quality by removing atmospheric blur from ground-based data.

The model runs on NVIDIA GPU-powered supercomputers and builds on prior Webb data-processing improvements, enabling faster, deeper analysis without replacing astronomers and helping with quicker discovery and pattern detection.

Rubin Observatory began science operations in 2025 and will generate about 20 terabytes of data per night as part of the Legacy Survey of Space and Time, creating a substantial data-handling challenge that Neo addresses.

Previous machine-learning tools for JWST sped up data processing, enabling earlier identification of mature galaxies.

A new neural network named Neo, a conditional generative adversarial network, is trained on matched image pairs from Subaru and the Hubble Space Telescope to reconstruct fine details lost to atmospheric distortion by comparing ground-based and space-based views of the same targets.

UC Santa Cruz researchers developed Neo to remove atmospheric blur from ground-based observations and enhance image resolution for the Vera C. Rubin Observatory in Chile.

AI image processing drastically speeds up JWST data analysis, reducing timelines from years to days or less and accelerating potential discoveries.

Neo runs on GPU-accelerated systems and classifies image pixels into objects like stars, disc galaxies, or empty sky, illustrating that AI is intended to augment rather than replace astronomers by handling vast data, flagging unusual objects, and revealing subtle patterns.