NASA and IBM have collaborated to develop and release the NASA-IBM Lunar Foundation Model, an open-source AI system. This model is now available for download on Hugging Face, providing researchers with a new tool for lunar exploration and analysis.
The initiative supports NASA's ongoing Artemis missions, which aim to return humans to the Moon and establish a sustained presence.
The Lunar Foundation Model has shown particular aptitude in identifying areas on the lunar surface that may contain ice. In tests, the model reduced errors by 23 percent when compared against SwinV2-B, a Microsoft-trained vision system often used as a baseline for image analysis.
Additionally, the model outperformed SwinV2-B by 19 percent in identifying and classifying craters, while utilizing half the amount of training data. Its capabilities were verified when it correctly identified a new crater formed by a SpaceX Falcon 9 rocket impact.
Training the model presented challenges due to the unique lighting conditions on the Moon. Unlike Earth observation, where atmospheric scattering softens shadows, lunar shadows are sharp and completely dark, meaning shadowed pixels carry no light information. The model was developed to handle these distinct imaging characteristics.
✨ This summary was generated by AI from the outlets' reporting listed below. It is not independently verified and may contain errors — check the original sources. How BrevFeed works →
One email each morning: the day's tech stories, clustered across outlets and summarized. No account needed.
One email a day. Unsubscribe in one click, any time.
Spend a few minutes, get the whole day. Every topic's top stories in one hands-free rundown — listen, watch, or read the transcript.
▶ Play today's briefNew every morning, and the back catalogue is archived by date.
NASA and IBM released the open-source NASA-IBM Lunar Foundation Model on Hugging Face, an AI system designed for lunar exploration. The model demonstrated improved accuracy in identifying lunar ice and classifying craters compared to existing vision systems, using less training data.