sentinel-hub/eo-learn

What it solves

It simplifies the process of extracting valuable information from satellite imagery. It addresses the challenge of processing massive amounts of high-resolution spatio-temporal data from programs like Copernicus and Landsat, making it easier to automatically detect complex patterns for applications like land use monitoring, disaster control, and humanitarian relief.

How it works

eo-learn acts as a bridge between the Earth observation field and the Python data science ecosystem. It uses NumPy arrays to handle remote sensing data and allows users to define a sequence of operations (a processing chain) to extract information. The library is modular, consisting of several specialized modules:

• Core: Implements basic building blocks like EOPatch, EOTask, and EOWorkflow.
• Coregistration: Handles image co-registration.
• Features: Provides utilities for feature manipulation and data property extraction.
• Geometry: Manages transformations between vector and raster data.
• IO: Manages data acquisition from Sentinel Hub services or local storage.
• Mask: Calculates masks for clouds, snow, and other obstructions.
• ML-tools: Provides tools for pre- and post-processing in machine learning workflows.
• Visualization: Offers tools to visualize core elements.

Who it’s for

It is designed for both remote sensing experts who want to use state-of-the-art computer vision and machine learning tools, and non-experts who want an easier entry point into the field of Earth observation.

Highlights

• Modular Design: Encourages the sharing and reuse of specific tasks like cloud masking and classification.
• Python Ecosystem Integration: Seamlessly connects satellite data to NumPy and other ML/DL libraries.
• Spatio-Temporal Focus: Specifically built to handle image sequences over time.
• Flexible Deployment: Can be installed via PyPI, Conda, or run as a pre-configured Docker image with Jupyter notebooks.