Home
One line of code to get embeddings from any Remote Sensing Foundation Model (RSFM) for any location and any time
Start Here¶
- Read Quickstart to install and run a first example
- Read Concepts to understand temporal/output semantics
- Use Workflows to pick the right API for your task
- Go to Model Overview for the comparison matrix, preprocessing notes, and temporal behavior
- Open Model Reference when you need advanced details and exact assumptions
- Start with API Reference
- Then use Specs & Data Structures for
TemporalSpec,OutputSpec, and related types
Suggested reading path
If you're unsure where to start, use Quickstart → Concepts → Workflows.
Common Tasks¶
| Goal | Best Entry Point | Main API |
|---|---|---|
| Get one embedding for one ROI | Quick Start | get_embedding(...) |
| Compute embeddings for many ROIs (same model) | Common Workflows | get_embeddings_batch(...) |
| Build an export dataset for experiments | Common Workflows | export_batch(...) |
| Debug bad inputs/clouds/band issues | Common Workflows | inspect_provider_patch(...) (recommended) |
| Compare model preprocessing and I/O assumptions | Supported Models | model matrix + notes |
Motivation¶
The remote sensing community has seen an explosion of foundation models in recent years. Yet, using them in practice remains surprisingly painful: * Inconsistent model interfaces (imagery vs. tile embeddings) * Ambiguous input semantics (patch / tile / grid / pooled) * Large differences in temporal, spectral, and spatial requirements * No easy way to fairly compare multiple models in a single experiment
RS-Embed aims to fix this.
Goal
Provide a minimal, unified, and stable API that turns diverse RS foundation models into a simple ROI → embedding service — so researchers can focus on downstream tasks, benchmarking, and analysis, not glue code.
Why rs-embed?¶
- Unified interface for diverse embedding models (on-the-fly models and precomputed products).
- Spatial + temporal specs to describe what you want, not how to fetch it.
- Batch export as a first-class workflow via
export_batch. - Compatibility wrappers preserved (for example
export_npz,inspect_gee_patch) without changing the main learning path.
Documentation Map¶
Learn¶
- Quickstart: installation + first successful runs
- Concepts: mental model (
TemporalSpec,OutputSpec, andbackend="auto"access routing)
Guides¶
- Workflows: task-oriented usage patterns
- Model Overview: model capabilities, preprocessing, env knobs
Reference¶
- API Reference: exact signatures and parameter details
- Limitations: current constraints and known edge cases
Development¶
- Extending: add new model adapters and integrate with registry/export