geospaitial-lab-aviary 1.0.0b4

Composable inference and postprocessing pipeline for remote sensing data

aviary provides composable components for tile-based processing of geospatial data. This enables you to easily run models on large datasets, export the predictions in a georeferenced file format and postprocess them for further downstream tasks.
Besides the pipelines, aviary also provides task-specific models for remote sensing applications.

aviary is designed upon the following concepts:

• High-level Python API
  Abstract components for building pipelines without boilerplate code

• Command-line interface (CLI)
  Run the pre-built pipelines easily without writing any code

• Customizable pipelines
  Compose your own pipelines with the provided components

• Extensible components
  Add your own components to the pipeline

• Support for large datasets
  Tile-based processing for large datasets (local, remote or web services)

• Support for geospatial data
  Export predictions as geodata, ready for downstream tasks

• Task-agnostic
  Process geospatial data with a range of machine learning tasks

• ML-framework agnostic
  Use your favorite machine learning framework

Installation

You can choose between two installation methods, whether you need access to the Python API or the command-line interface (CLI) only. If you just want to use the pre-built pipelines with the command-line interface, you can use the Docker image.

Installation with pip

⚠️ Note: aviary is currently released as a pre-release version. To install the latest version, you need to add the --pre flag.

pip install geospaitial-lab-aviary

Note that aviary requires Python 3.10 or later.

Have a look at the installation guide for further information.

Installation with uv

uv pip install geospaitial-lab-aviary

Note that aviary requires Python 3.10 or later.

Have a look at the installation guide for further information.

Installation with Docker

docker pull ghcr.io/geospaitial-lab/aviary

Have a look at the installation guide for further information.

Next steps

Have a look at the how-to guides to get started.

Documentation

The full documentation is available at geospaitial-lab.github.io/aviary.

About

aviary is developed by the geospaitial lab at the Westfälische Hochschule - Westphalian University of Applied Sciences in Gelsenkirchen, Germany.