The Allen Institute for AI’s OlmoEarth Studio now lets users compute and export embedding vectors from its open-source Earth-observation foundation models. Three encoder variants — Nano (128-dim), Tiny (192-dim), and Base (768-dim) — produce embeddings stored as signed 8-bit integers in Cloud-Optimized GeoTIFF format. These embeddings support downstream applications such as similarity search, few-shot classification with minimal labeled data, temporal change detection, and unsupervised landscape analysis via dimensionality reduction techniques like PCA.
