LunarFM
A multimodal foundation model for lunar science. Trained on orbital imagery, elevation, gravity, and spectral data to enable intelligent exploration of the Moon.
Paper
Read the preprint to learn more about how we put together this first version of LunarFM.
Science Release
~200k moon-surface embeddings, 2000 sample chips (Apollo 11 site), downstream-task datasets, code and model weights, and usage notebooks. Download the data.
Tutorials
Get started quickly with our tutorial notebooks covering inference, fine-tuning, and data loading. Browse the code.
🚀 To access the science release and the tutorials, please fill out this form.