Stocks
Moon's Vast Image Archive Gets AI-Boosted Search Capabilities
NASA and IBM have collaborated to create an artificial-intelligence model for scientists to search the Moon's vast image archive.
The system, trained on roughly two million lunar image tiles from the Lunar Reconnaissance Orbiter, aims to make existing observations more useful by allowing researchers to adapt a common model to particular research questions instead of starting each task from scratch.
The team's technical description explains that lighting geometry is supplied explicitly alongside the observations, and different kinds of imagery and terrain information are combined in training, rather than relying on visible appearance alone.