Skip to content

umbra-py

A Python-first toolkit to make Umbra SAR open data easy to discover, load, download, and analyze.

Umbra publishes very-high-resolution (down to ~16–25 cm) synthetic aperture radar (SAR) imagery as open data under a permissive CC BY 4.0 license. The data is excellent, but getting started is hard: it ships in specialized formats (SICD, SIDD, CPHD, GEC), is indexed by a large static STAC catalog, and the existing tooling is low-level. umbra-py makes working with it feel as approachable as working with Sentinel-1 or Landsat.

Try it now — no install

Browse and search the whole open archive from your browser in the live showcase: a zoomable whole-catalog map and an interactive explorer, hosted on GitHub Pages, no account or download required. Build your own with umbra showcase.

Status

v0.1 / early alpha. The discovery + download core works against Umbra's live catalog today; processing helpers are minimal and growing.

What it gives you

  • Discovery — search Umbra's 17+ TB static STAC catalog by area, date, and product type, by place name, or by polygon — no search API required. Build a local SQLite index (or fetch the prebuilt weekly snapshot) for near-instant offline repeats.
  • Download — resume-safe HTTPS downloads with integrity verification.
  • Convert — turn a complex SICD product into a map-ready, geocoded cloud-optimized GeoTIFF (umbra convert), with optional DEM terrain orthorectification, geoid correction, and radiometric terrain flattening.
  • Load — read a clipped/decimated scene straight into xarray.
  • Visualize — interactive Folium maps, HTML thumbnail galleries, full-res browser viewers, before/after swipes, change composites, and time scans.
  • ML prep — cut scenes into georeferenced training chips with metadata.
  • AI-native surfaces — an MCP server (umbra-mcp), a read-only STAC API (umbra serve), native LangChain / LlamaIndex tools, plus model-backed natural-language search, scene description, and visual similarity.

Next steps

License & attribution

umbra-py is Apache-2.0. Umbra's open imagery is licensed CC BY 4.0 — attribute "Umbra Lab, Inc." when you publish derived products. This is an independent, unofficial toolkit and is not affiliated with or endorsed by Umbra Lab, Inc.