NASA-IBM Lunar Foundation Model
Analyzes lunar orbital imagery for craters, irregular mare patches and polar ice prospectivity.
- Field
- Planetary science
- Method
- Foundation model
- Data it takes
- imagery
- License
- Apache-2.0
This is a vision transformer pretrained on lunar remote-sensing data across 11 modalities, released with fine-tuned models for three tasks: crater detection in LROC narrow- and wide-angle camera imagery, segmentation of irregular mare patches, and a map of polar ice prospectivity built from thermophysical and illumination layers.
The ice-prospectivity output reproduces an expert’s map of where ice is plausible. It does not measure ice. The model also carries no geodetic reference frame, so locations come from the input data rather than from the model.
It is a different model from LunarFM, despite the similar name. The code, backbone and all released fine-tunes are Apache-2.0.
Catalog entry last checked 2026-10-03. All models →
Run NASA-IBM Lunar Foundation Model on your data
The lab runs this on its own compute, in a container, with the inputs and parameters recorded alongside the result. Initial scoping conversations are free.
Contact the Lab