NASA · Space

NASA and IBM Launch AI Model for Lunar Science

NASA and IBM have launched an open-source AI foundation model to analyze lunar data, trained primarily on data from NASA's Lunar Reconnaissance Orbiter.

The NASA-IBM Lunar Foundation Model is available on Hugging Face and GitHub, aiming to help researchers analyze vast lunar surface data for geological understanding and future exploration planning.

Foundation models are pre-trained on large datasets, enabling versatile and efficient scientific research through quick fine-tuning for specific tasks.

The model was trained on millions of image tiles from LRO, along with data from other missions like GRAIL and JAXA's Selenological and Engineering Explorer.

Scientists can adapt the model to map craters, identify volcanic features, and estimate polar ice stability with minimal labeled data.

It aids in studying lunar volcanism by identifying irregular mare patches and efficiently mapping craters, crucial for dating the lunar surface.

The model performed comparably or better than other models, especially in estimating polar ice stability, as part of NASA's AI for science strategy.

This open-science release includes datasets and benchmarks to support reproducible research, though lighting variations may affect small crater visibility.

AI-samenvatting op basis van de bron.