tessera

Download TESSERA satellite embeddings for a specific bounding box by specifying the region and year to fetch dense mosaics as an (H, W, 128) float32 array. Export results directly to GeoTIFF, NPY, or Zarr formats.

Sample satellite embeddings at specific latitude and longitude coordinates to retrieve an (N, 128) float32 array. Missing points are automatically filled with NaNs, making it ideal for labeled site extraction and sparse classification tasks.

Yes, you can use the geotessera CLI, Python, or R libraries to export geospatial ML embeddings to Zarr. The process supports configurable bands, hash-verified cached registry access, and optional output formats including GeoTIFF and NPY.

Yes, coverage checking is supported before fetching TESSERA tiles. You can verify available data for a specified year and region file, ensuring hash-verified cached registry access prevents downloading unnecessary or missing tiles.

Available output formats for dense mosaics from satellite embedding tiles include GeoTIFF, NPY, and Zarr. These formats support wall-to-wall raster analysis like spatial clustering and land cover classification with configurable bands.

Point samples from TESSERA embeddings contain NaN values to represent missing data at requested (lon, lat) locations. This ensures the output (N, 128) float32 array maintains consistent dimensions for downstream sparse classification and labeled site extraction.