BACK_TO_FEEDAICRIER_2
rs-embed unifies remote sensing foundation model embeddings
OPEN_SOURCE ↗
REDDIT · REDDIT// 8d agoOPENSOURCE RELEASE

rs-embed unifies remote sensing foundation model embeddings

rs-embed is a Python library that provides a unified, "one-line" interface for obtaining embeddings from various Remote Sensing Foundation Models (RSFMs) for any location and time. By abstracting the complexities of satellite data fetching via Google Earth Engine and preprocessing through TorchGeo, it allows researchers and developers to task models with the same ease as tasking satellites.

// ANALYSIS

Standardizing remote sensing workflows is a massive leap for reproducible research in earth observation.

  • Abstracts fragmented model-specific "glue code" into a clean, ROI-centric API
  • Supports major models including IBM/NASA's Prithvi, SatMAE, RemoteCLIP, and AnySat
  • Built-in orchestration handles prefetching, inference, and large-scale batch exports
  • Integrates both on-the-fly inference and access to massive precomputed datasets like Tessera
  • Open-source release (Apache-2.0) from the CyberGIS Center at UIUC
// TAGS
rs-embedembeddingdata-toolsopen-sourceresearchmultimodal

DISCOVERED

8d ago

2026-04-03

PUBLISHED

8d ago

2026-04-03

RELEVANCE

7/ 10

AUTHOR

amritk110