gaitmap 2.6.0

The Gait and Movement Analysis Package - Your definite guide to reliable IMU based human movement analysis.

gaitmap - The Gait and Movement Analysis Package

gaitmap provides a set of algorithms to analyze your IMU movement data (with a focus on foot-worn IMUs) without getting into your way.

• 💻 20+ Algorithms from 17+ publications
• 📚 Extensive documentation
• 📝 Build to be easily extendable
• ⚙️ Familiar API inspired by scikit-learn
• 🤝 Interoperable with the other libraries from the gaitmap ecosystem (gaitmap-dataset, tpcp, ...)

Documentation: gaitmap.readthedocs.io
Learn more about the gaitmap ecosystem: the gaitmap ecosystem

Installation

Gaitmap is split into two packages: gaitmap and gaitmap_mad (Learn more). To get access to all available algorithms, you need to install both packages:

pip install gaitmap gaitmap_mad --upgrade

Both packages are always released together and have the same version number. We don't recommend mixing different versions of gaitmap and gaitmap_mad.

Note: gaitmap-mad is published under a AGPL-3.0 license, while gaitmap is published under a MIT license. Please, check the License section below for more details.

In case you are sure that AGPL-3.0 is compatible with your project, you can install gaitmap_mad without any downsides. Otherwise, just install gaitmap and check the API-docu page of the individual algorithms if they are available in gaitmap only.

Installing from Github

If you want to install the latest version from Github, you can use the following command:

# For gaitmap
pip install "git+https://github.com/mad-lab-fau/gaitmap.git" --upgrade
# For gaitmap_mad
pip install "git+https://github.com/mad-lab-fau/gaitmap.git#subdirectory=packages/gaitmap_mad"

Note, that we don't guarantee that the latest version on Github is stable.

Enabling specific features

Hidden Markov Models

To use the HMM (anything imported from gaitmap.stride_segmentation.hmm) based algorithms make sure you install gaitmap with the hmm extra.

pip install gaitmap_mad "gaitmap[hmm]" --upgrade

This installs the pomegranate package, which is the basis for the HMM implementation. Note, that we only support the pomegranate version >=0.14.2,<=0.14.6 and that pomegrante is not compatible with Python 3.10.

We are working on upgrading to a newer version of pomegranate, but this is not a priority at the moment. You can track the progress in the pull request.

Supported Python versions

gaitmap is tested against Python 3.9 at the moment. We expect most features to work with all Python versions >= 3.9, but because of some known issues (see specific features above) we do not officially support them.

Working with Algorithms

gaitmap is designed to be a toolbox and not a single algorithm. This means, that you are expected to pick and use individual algorithms.

To get started with gaitmap we recommend to follow these steps:

1. Have a look at our example to get an understanding of what gaitmap can do: Example
2. Understand the common datatypes we use: Common Datatypes, Coordinate Systems
3. Learn how to prepare your own data (or play around with example data): Prepare Data
4. Check the API docs and examples for available algorithms: API Docs, Examples

Each algorithm is implemented as a class, which is initialized with the required parameters. In many cases the default parameters are sufficient, but to get the best results on your data, you should adapt them.

from gaitmap.stride_segmentation import BarthDtw

stride_segmenter = BarthDtw(max_cost=2.5)

After initialization, you can apply the algorithm to your data (each group of algorithm has a different action method):

my_imu_data = ...
stride_segmenter.segment(my_imu_data, sampling_rate_hz=204.8)

Results are stored as parameters of the algorithm object (with a trailing underscore):

segmented_strides = stride_segmenter.stride_list_

Most algorithms have additional results available to provide further information.

dtw_warping_path = stride_segmenter.paths_

To build a full gait analysis pipeline you can combine multiple algorithms (Example). You can even substitute your own algorithms (Guide) or use the provided tooling to validate and optimize your algorithms using tpcp (General Guide, Example).

Contributing

We want to hear from you (and we want your algorithms)!

👍 We are always happy to receive feedback and contributions. If you run into any issues or have any questions, please open an issue on Github or start a discussions.

📚 If you are using gaitmap in your research or project, we would love to hear about it and link your work here! The show and tell section in the discussions is a great place for this.

💻 And most importantly, we want your algorithms! If you have an algorithm that you think would be a good fit for gaitmap, open an issue, and we can discuss how to integrate it. We are happy to help you with the integration process. Even if you are not confident in your Python skills, we can discuss ways to get your algorithm into gaitmap.

For all these topics check out our contributing guide for more details.

Citation

If you use gaitmap in your research we would appreciate a citation. This helps us to justify the time we invest in the development and maintenance of the library.

A. Küderle et al., "Gaitmap—An Open Ecosystem for IMU-Based Human Gait Analysis and Algorithm Benchmarking," in IEEE Open Journal of Engineering in Medicine and Biology, vol. 5, pp. 163-172, 2024, doi: 10.1109/OJEMB.2024.3356791.

@ARTICLE{10411039,
  author={Küderle, Arne and Ullrich, Martin and Roth, Nils and Ollenschläger, Malte and Ibrahim, Alzhraa A. and Moradi, Hamid and Richer, Robert and Seifer, Ann-Kristin and Zürl, Matthias and Sîmpetru, Raul C. and Herzer, Liv and Prossel, Dominik and Kluge, Felix and Eskofier, Bjoern M.},
  journal={IEEE Open Journal of Engineering in Medicine and Biology}, 
  title={Gaitmap—An Open Ecosystem for IMU-Based Human Gait Analysis and Algorithm Benchmarking}, 
  year={2024},
  volume={5},
  number={},
  pages={163-172},
  keywords={Pipelines;Benchmark testing;Software algorithms;Ecosystems;Machine learning algorithms;Estimation;Trajectory;Accelerometer;walking;biomarker;biomechanics;movement analysis},
  doi={10.1109/OJEMB.2024.3356791}}

If you use a specific algorithm please also make sure you cite the original paper of the algorithm! We recommend the following citation style:

We used the algorithm proposed by Author et al. [paper-citation], implemented by the Gaitmap package [gaitmap-citation].

License

The gaitmap (which includes most algorithms) is published under a MIT license. This is a permissive license, which allows you to use the code in nearly any way you want, as long as you include the original license in you modified version.

gaitmap_mad is published under a AGPL-3.0 license. This only affects algorithms that were developed by the MaD Lab in collaboration with industry partners. The AGPL-3.0 license is a copyleft license, which if you integrate gaitmap_mad into your software and provide it to others (either as a service or as a product), you have to publish your software under a compatible open source license. Please, check the AGPL-3.0 license for more details and make sure you understand the implications.

If you need to use gaitmap_mad in a closed source project, please contact us for a potential commercial license.

For Developers

The Development Guides have detailed information for all new developers.

Here some quick references Install Python >=3.9 and uv. Then run the commands below to get the latest source and install the dependencies:

git clone https://github.com/mad-lab-fau/gaitmap.git
# For Python 3.9 (and if you need to work on hmm)
uv sync --group dev --all-extras
# For Python >=3.10 (you can not work on hmm stuff with this)
uv sync --group dev --extra stats

Warning: Building the docs currently only works in 3.9 with all extras installed.

Note, that you don't need to care about the gaitmap_mad subpackage. All dependencies are specified in the main pyproject.toml and the gaitmap_mad will be installed in editable mode when running uv sync --group dev ....

To run any of the tools required for the development workflow, use the poe commands:

uv run poe
...
CONFIGURED TASKS
  format            
  lint              Lint all files with ruff.
  ci_check          Check all potential format and linting issues.
  test              Run Pytest with coverage.
  docs              Build the html docs using Sphinx.
  register_jupyter  Register the gaitmap environment as a Jupyter kernel for testing.
  version           Bump version in all relevant places.
  bump_dev          Update all dev dependencies to their @latest version.