This is the Salzburg Database of Geometric Inputs, a repository of graphs for testing and benchmarking your computational geometry implementations.
All data can be found in the db directory. It can be downloaded either
directly or with git-annex using
git clone https://sbgdb.cs.sbg.ac.at/db/.git
followed by git annex get
in the
working copy directory. Git annex enables you to also get specific files only.
At the moment, our repository consists of random simple polygons, with and without holes, produced with different generators. Furthermore, we have instances of specific polygons such as the Koch snowflake.
Releases
The latest release of the data is sbgdb-20200507
,
released on May 7, 2020. This data-set is fixed and will not see updates.
It can also be retrieved from zenodo; its persistent identifier
is doi:10.5281/zenodo.3784788
.
Future updates will happen in the work-in-progress directory wip
,
which started as a copy of the previous release. In time, this data-set will
become the next release.
Citing this dataset
If you use data or tools provided by this database, we would appreciate a short email to held@cs.sbg.ac.at with a description of how you use our data or tools. This aids us in estimating what parts are used and useful and will help guide the direction of this project.
In any case, if you use this database while working on a publication or to get some other real work done, please:
- provide the URL of this database, and, if applicable,
- cite the following paper: Salzburg Database of Polygonal Data: Polygons and Their Generators; Günther Eder, Martin Held, Steinþór Jasonarson, Philipp Mayer, and Peter Palfrader; Data in Brief, 31:105984, August 2020; doi:10.1016/j.dib.2020.105984.
We also provide BibTeX snippets.
Formats
The graphs in this repository are provided in xz-compressed graphml with additional attributes for storing the vertex coordinates of the embedding as well as for instance multiplicative edge weights.
Most graphs are edge weighted with random weights. This is for users who would like to use and test their code with weighted input. Users are of course free to ignore the weights and assume unit weights if it makes sense for their use-case.