GeoAI SDK Documentation

GeoAI SDK Documentation#

Python SDK for running geospatial AI models at scale on Azure AI Foundry and Planetary Computer.

Getting Started#

  1. Install: See Installation below

  2. Learn: Try the EOOS notebook or MARS notebook

  3. Reference: Use API docs below for parameter details

API Reference#

Core Classes#

  • Input - Configure data sources (STAC catalogs, collections)

  • Output - Configure result destinations (GeoCatalog, blob storage)

  • Constraint - Filter by space, time, and properties

  • Models - Available models (EOOS, MARS)

Features#

  • Validation - Pre-flight checks for input, model, and output

Examples#

All examples are available as Jupyter notebooks in examples/:

  • EOOS Object Detection - Detect objects in satellite imagery

  • MARS Map Generation - Extract buildings, roads, railways, water

Installation#

From Wheel#

Download the wheel from the GitHub Release:

pip install geoai_sdk-0.1.0-py3-none-any.whl

From Cloned Repo#

pip install tools/geoai-sdk/

With Notebook Dependencies#

# From wheel
pip install "geoai_sdk-0.1.0-py3-none-any.whl[examples]"

# From cloned repo
pip install "tools/geoai-sdk[examples]"

Development#

# For contributors: install with dev tools
pip install "tools/geoai-sdk[dev]"

# Format code before commits
python -m black geoai/ --line-length 100
python -m isort geoai/ --profile black --line-length 100
python -m flake8 geoai/ --count --select=E9,F63,F7,F82

Architecture#

The SDK is built with three layers:

1. Public API Layer#

User-facing classes for configuration:

2. Executors Layer#

Handles workflow execution:

  • LocalExecutor - Sequential processing with progress bars

3. Core Components Layer#

Reusable business logic:

  • Spatial - ChipMaker, DetectionMerger

  • STAC - STACSearcher, ItemSelector

  • Imagery - ImageFetcher, ImagePreprocessor

  • Models - ModelClient, PayloadBuilder, ResponseParser

  • Results - DetectionFormatter, GeoCatalogWriter

Supported Models#

Model

Type

Input

Output

EOOS

Object Detection

NAIP

GeoJSON Features

MARS

Map Generation

NAIP

GeoJSON Features + PNG

Support#

  • Team: Azure Orbital Spatio Team

License#

This project is licensed under the MIT License - see the LICENSE file for details.

Contents