Revalidation#

This page shows the most up-to-date results on the validation of all algorithms implemented in mobgap. The validation is performed based on the Mobilise-D TVS dataset, which can be downloaded from Zenodo.

As most mobgap algorithms are re-implementations of algorithms that were originally available only in Matlab, the revalidation contains comparisons with these original implementations. For each algorithm/algorithmic block, two types of subpages are available on this page: First, an overview of the analytic results and second, a document that contains the code to reproduce these results. Both files are computational notebooks that can be downloaded either from Github or by using the available download button on the respective pages.

In case you are interested in performing additional analysis, we recommend starting by downloading these notebooks as a starting point. The notebooks will automatically download the required files containing the results from the result repository and are mostly self-contained and only require mobgap to be installed to run them. They are further written in a way that all relevant code is directly available in the notebook and not deeply hidden in some library, which should allow easy modification. The code available to generate the results can also be used as a starting point to run the same validation on any other dataset.

These results will be regularly updated when there are major updates to the algorithms and new releases of the TVS dataset. The current results are generated using the TVS dataset version 1.0.2 (released at 2025-07-11).

To view older versions of this documentation and results, use the version selector in the bottom right of this page.