Landmarker is a PyTorch-based toolkit for (anatomical) landmark localization in 2D/3D images. It is designed to be easy to use and to provide a flexible framework for state-of-the-art landmark localization algorithms for small and large datasets. Landmarker was developed for landmark detection in medical images. However, it can be used for any type of landmark localization problem.
| command | |
|---|---|
| pip | pip install landmarker |
Technical documentation is available at documentation.
Examples and tutorials are available at examples
- Modular: Landmarker is designed to be modular. Almost all components can be used independently.
- Flexible: Landmarker provides a flexible framework for landmark detection, allowing you to easily customize your model, loss function, and data loaders.
- State-of-the-art: Landmarker provides state-of-the-art landmark detection models and loss functions.
- Extension to landmark detection in videos.
- ...
We welcome contributions to Landmarker. Please read the contributing guidelines for more information.
If you use landmarker in your research, please cite the following paper:
J. Jonkers, L. Duchateau, G. Van Wallendael, and S. Van Hoecke, “landmarker: A Toolkit for Anatomical Landmark Localization in 2D/3D Images,” SoftwareX, vol. 30, p. 102165, May 2025, doi: 10.1016/j.softx.2025.102165.
J. Jonkers, F. Coopman, L. Duchateau, G. V. Wallendael, and S. V. Hoecke, “Reliable uncertainty quantification for 2D/3D anatomical landmark localization using multi-output conformal prediction,” Mar. 18, 2025, arXiv: arXiv:2503.14106. doi: 10.48550/arXiv.2503.14106.
Landmark is licensed under the MIT license.
👤 Jef Jonkers