AI2's OlmoEarth Studio Adds Custom Embedding Exports for Earth Observation Analysis
Key Takeaways
- ▸OlmoEarth Studio now enables on-demand custom embedding generation with configurable parameters for time, resolution, and data sources
- ▸Three encoder variants (Nano 128-dim, Tiny 192-dim, Base 768-dim) balance performance and computational cost for different use cases
- ▸Open-source foundation models and published research paper allow independent verification and custom embedding computation by the community
Summary
AI2 has expanded OlmoEarth Studio with the ability to compute and export custom embedding vectors from satellite imagery. The embeddings provide compact numerical representations of Earth-observation data that can be applied to downstream tasks including similarity search, segmentation, and unsupervised exploration. Users can now configure embeddings by selecting their area of interest, time range, encoder variant (Nano, Tiny, or Base), spatial resolution (10-80 meters per pixel), and imagery sources through the Studio UI or API, with results delivered as lightweight Cloud-Optimized GeoTIFFs (COGs). The OlmoEarth foundation models and complete source code are open-source and publicly available, allowing the research community to inspect and replicate the embedding generation process.
