NASA and IBM Unveil Lunar Mapping AI Model

NASA and IBM have developed an AI model to improve the mapping of the Moon. The model, called the Lunar Foundation Model, was trained on about two million image patches from various NASA missions, including the Lunar Reconnaissance Orbiter (LRO) and GRAIL.

The dataset brings together more than thirty geographically aligned layers, coming from nine instruments and four missions, providing different information about the same terrain. This includes views as low as one meter per pixel and larger areas across multiple light bands at a resolution of one hundred meters per pixel.

Researchers will be able to adapt this base to their own work by feeding it already identified examples. The model is being used for various tasks, including crater counting, spotting small volcanic formations, and searching for ice at the poles.