austriadownloader

• Github repository: https://github.com/Zerhigh/austriadownloader/
• Documentation https://Zerhigh.github.io/austriadownloader/

Introduction

This repository contains the code for developing and testing the austriadownloader package, capable of downloading temporally and spatially aligned Austrian Orthophoto (RGB & NIR) and Cadastral data for (among others) Deep Learning application. Available datasets include the Austrian Orthophoto series of 2024 (includes image tiles from 2021 to 2023) and the corresponding Cadastral datasets (eg. from 01.04.2021) published bi-anually. For a detailed analysis and description of datasources, processing steps, and methodology please refer to the corresponding publication (to be added upon release, contact authors for a pre-print until then).

Getting Started

All required meta-datasets are available in austriadownloader/austria_data/ and can be created by executing austriadownloader/austria_data/metadata_creation.py.

Provide sample image POIs as centroids in a dataframe with the following scheme in the WGS84 CRS (EPSG:4326). Image dimensions will be determined by other input parameters such as pixel_size and shape. An independent (but closely related) git-repository for automatically creating such a sample file is available under austriadownloader_sampler.

Sample file structure:

An example for a sample file:

Code Example:

Refer to demo/demo.py for code, config, and sample files.

from pathlib import Path
from austriadownloader.downloadmanager import DownloadManager
from austriadownloader.configmanager import ConfigManager

config_path = Path("path_to_your_config.yml")
manager = DownloadManager(config=ConfigManager.from_config_file(config_path))
manager.start_download()

Input parameters are provided in the config file and include:

Available Classes

To select your class labels, select one or more from the following list (Source: BEV, page 12 ff.):

Results

Multi-label mask with all available cadastral classes selected (not all are present in the selected sample):

General overview of different cadastral classes merged into a binary mask:

Selection of unique cadastral classes:

Citation

This repository was created for a presentation at the AGIT 2025 conference.

Repository initiated with fpgmaas/cookiecutter-poetry.