SourceIntroducing OlmoEarth embeddings: Custom embedding exports from OlmoEarth Studio for downstream analysis
This allows users to perform geospatial tasks like land-cover mapping or change detection more efficiently using far fewer labeled pixels than traditional methods.
- Users can compute and export embedding vectors generated by open-source OlmoEarth foundation models
- Supports multiple encoder variants including Nano (1.4M parameters), Tiny (6.2M parameters), and Base (89M parameters)
- Outputs Cloud-Optimized GeoTIFFs (COGs) with one band per embedding dimension encoded as signed 8-bit integers
- Supports downstream tasks including similarity search, few-shot segmentation, change detection, and unsupervised PCA exploration