IBM and NASA Open-Source an AI Model for Lunar Exploration

IBM and NASA released the open-source NASA-IBM Lunar Foundation Model for lunar science exploration. The model improves identification of key surface geographic features by up to 23% compared with commonly used methods. When identifying potential lunar ice regions, it reduces RMSE by up to 22% compared with SwinV2-B. It also comes with the first open-source lunar dataset, which brings together 4 missions, 9 instruments, and more than 30 spatially aligned layers.