Release of open AI model trained on 17 years of lunar data maps ice, craters and volcanoes

A NASA-IBM open AI model trained on 17 years of lunar data has been released, mapping ice, craters, and volcanoes. The model, called the NASA-IBM Lunar Foundation Model, was trained on 2 million image tiles from NASA's Lunar Reconnaissance Orbiter (LRO) data, including over 1 million high-resolution images and nearly 964,000 multispectral images. The model supports researchers in analyzing lunar surface data efficiently, aiding in crater mapping, ice stability estimation, and geological studies. The model's performance matched or exceeded other baseline models in key tasks, showing potential for future lunar exploration.

The NASA-IBM Lunar Foundation Model is part of NASA's strategy to use AI for science, collaborating with IBM and several academic institutions. The model was developed by a team including experts from NASA, industry, and academia, and is hosted on Hugging Face for public access. It can help researchers identify and measure craters, estimate ice stability, and map volcanic features. The model's pre-training on LRO data allows it to adapt to various lunar research tasks with minimal labeled data. The model's release supports the global research community and includes pre-training datasets and benchmark collections. The model's success highlights the potential of AI in transforming lunar science and exploration.

The model's release is supported by a companion paper and is part of a growing collection of AI models developed through the NASA-IBM partnership. The model is integrated into the TerraTorch toolkit, and its open-source nature encourages scientific collaboration and reproducible research. The model's ability to map surface features and estimate ice stability makes it a valuable tool for understanding the moon's geological history and planning future exploration. The collaboration between NASA, IBM, and academic institutions demonstrates the potential of AI in advancing planetary science and space exploration.

AI-generated summary of the original article — see the source for the full story.