Random-forest automated bud annotation

    View project data

    License BSD-3 PyPI Python Version tests codecov napari hub

    Random-forest automated bud annotation

    This napari plugin was generated with Cookiecutter using @napari's cookiecutter-napari-plugin template.


    make sure you already have installed napari.

    Next, You can install napari-buds via pip:

    pip install napari-buds

    To install latest development version :

    pip install git+


    Contributions are very welcome. Tests can be run with tox, please ensure the coverage at least stays the same before you submit a pull request.


    Napari-Buds is a random forest based mother-bud annotation plugin for Napari devevoped by the TutucciLab ( of the systems biology group at the Vrije Universiteit van Amsterdam. Mother-bud annotation requires single or multichannel 2D images of budding yeast and some kind of marker that localizes to the bud. In the example dataset provided smFISH DNA-probes were used as localized bud marker.The GUI layout for random forest based classification was inspired by ImageJ 'plugin Weka Segmentation' [1].

    Please follow the workflow described underneath to perform mother-bud annotation:

    1. Open images in napari and create empty label layer. For multichannel images each channel should be provided seperately to napari. An example (jupyter) notebook for loading test data in napari is provided in the notebooks folder (named:Open Test Images Napari.ipynb ).

    2. If multichannel images are unaligned the translate widget under Plugins>napari-buds>Translate can be used. Select which layer should be translated to align to the layers in widget menu. Then use the aswd keys to translate (move) the selected layer. To register changes and update coordinates of the translated image in napari press t.

    Random forest classification

    1. To open the mother-bud annotation plugin go to Plugins>napari-buds>bud annotation.

    2. To train a random forest classifier, in the created label layer draw examples of cells, buds and background (see tutorial gif below). In the Define Label segment of the widget you define which label value (class #label_value) corresponds to cells, buds and background. Currently, cells and backgrounds and buds have to be defined in the Define Label segment if you want to be able to segment the classification aswell. In the segment Layers to extract Features from we can select which layers will be used in training the random forest classifier. Next press Train classifier. After training is completed a result layer is added to layer list. Inspect the results carefully to asses classifier performance. The trained classifier can be saved using the save classifier button. Previously trained classifier can be loaded by pressing Load classifier. Loaded classifer can applied to new images by pressing Classify, resulting again in a results layer. It is possible to change the random forest parameters with the Set random forest parameters button and changing the values in the pop up menu. Press Run to register changed settings. For an example of the parameters used see: and

    3. Next, we want to perfom watershed segmentation using the result layer. However, for watershed segmentation seeds (also called markers) are required (for an explanation of watershed segmenation see: To define the seeds we can either simply threshold on one of the supplied image layers or we can use distance tranform ( examples-segmentation-plot-watershed-py) if thresholding is not suitable.The resulting seeds layer can be adjusted manually by editing in napari. A good seeds layers correspond to each cell having a single seed (buds are not single cells). To perform watershed segmentation press the Segment button.

    4. Carefully inspect the resulting cell mask and bud layer. Correct the mistakes in both layers. Bud label values should correspond to the label value of the cell mask of mother cell. To verify mother bud relations were drawn correctly press Draw Mother-Bud relations. If Mother-Bud relations are correct, you can save both label layers. For an example of standardized saving and extracting the mother-bud relations from the saved layers see the 'napari-bud_example.ipynb' file in the notebooks folder.

    See video for clarification:

    Watch the video

    Similar Napari plugins

    1-napari-accelerated-pixel-and-object-classification (APOC) by Robert Haasse.



    Distributed under the terms of the BSD-3 license, "napari-buds" is free and open source software


    If you encounter any problems, please file an issue along with a detailed description.


    1. Arganda-Carreras, I., Kaynig, V., Rueden, C., Eliceiri, K. W., Schindelin, J., Cardona, A., & Sebastian Seung, H. (2017). Trainable Weka Segmentation: a machine learning tool for microscopy pixel classification. Bioinformatics, 33(15), 2424–2426. doi:10.1093/bioinformatics/btx180


    • 0.0.4

    Release date:

    • 20 August 2022

    First released:

    • 16 August 2022


    • BSD-3-Clause

    Supported data:

    • Information not submitted

    GitHub activity:

    • Stars: 0
    • Forks: 0
    • Issues + PRs: 0

    Python versions supported:

    Operating system:


    • numpy
    • magicgui
    • qtpy
    • pandas
    • napari
    • magic-class
    • scipy
    • sklearn
    • scikit-image
    • matplotlib
    • joblib

    Sign up to receive updates