Using the labels layer#

In this document, you will learn about the napari Labels layer, including using the layer to display the results of image segmentation analyses, and how to manually segment images using the paintbrush and fill buckets. You will also understand how to add a labels image and edit it from the GUI and from the console.

When to use the labels layer#

The labels layer allows you to take an array of integers and display each integer as a different random color, with the background color 0 rendered as transparent.

The Labels layer is therefore especially useful for segmentation tasks where each pixel is assigned to a different class, as occurs in semantic segmentation, or where pixels corresponding to different objects all get assigned the same label, as occurs in instance segmentation.

A simple example#

You can create a new viewer and add an labels image in one go using the napari.view_labels method, or if you already have an existing viewer, you can add a Labels image to it using viewer.add_labels. The api of both methods is the same. In these examples we’ll mainly use add_labels to overlay a Labels image onto on image.

In this example of instance segmentation, we will find and segment each of the coins in an image, assigning each one an integer label, and then overlay the results on the original image as follows:

import napari
from skimage import data
from skimage.filters import threshold_otsu
from skimage.segmentation import clear_border
from skimage.measure import label
from skimage.morphology import closing, square, remove_small_objects

coins = data.coins()[50:-50, 50:-50]
# apply threshold
thresh = threshold_otsu(coins)
bw = closing(coins > thresh, square(4))
# remove artifacts connected to image border
cleared = remove_small_objects(clear_border(bw), 20)
# label image regions
label_image = label(cleared)

# create the viewer and add the coins image
viewer = napari.view_image(coins, name='coins')
# add the labels
labels_layer = viewer.add_labels(label_image, name='segmentation')

Segmentation of coins in an image, displayed using a labels layer

Arguments of `view_labels` and `add_labels`#

view_labels() and add_labels() accept the same layer-creation parameters.

Show code cell content Hide code cell content

help(napari.view_labels)

Help on function view_labels in module napari.view_layers:

view_labels(data, *, affine=None, blending='translucent', cache=True, colormap=None, depiction='volume', experimental_clipping_planes=None, features=None, metadata=None, multiscale=None, name=None, opacity=0.7, plane=None, properties=None, rendering='iso_categorical', rotate=None, scale=None, shear=None, translate=None, visible=True, title='napari', ndisplay=2, order=(), axis_labels=(), show=True) -> napari.viewer.Viewer
    Create a viewer and add a labels layer.
    
    Parameters
    ----------
    data : array or list of array
        Labels data as an array or multiscale. Must be integer type or bools.
        Please note multiscale rendering is only supported in 2D. In 3D, only
        the lowest resolution scale is displayed.
    affine : n-D array or napari.utils.transforms.Affine
        (N+1, N+1) affine transformation matrix in homogeneous coordinates.
        The first (N, N) entries correspond to a linear transform and
        the final column is a length N translation vector and a 1 or a napari
        `Affine` transform object. Applied as an extra transform on top of the
        provided scale, rotate, and shear values.
    blending : str
        One of a list of preset blending modes that determines how RGB and
        alpha values of the layer visual get mixed. Allowed values are
        {'opaque', 'translucent', and 'additive'}.
    cache : bool
        Whether slices of out-of-core datasets should be cached upon retrieval.
        Currently, this only applies to dask arrays.
    colormap : CyclicLabelColormap or DirectLabelColormap or None
        Colormap to use for the labels. If None, a random colormap will be
        used.
    depiction : str
        3D Depiction mode. Must be one of {'volume', 'plane'}.
        The default value is 'volume'.
    experimental_clipping_planes : list of dicts, list of ClippingPlane, or ClippingPlaneList
        Each dict defines a clipping plane in 3D in data coordinates.
        Valid dictionary keys are {'position', 'normal', and 'enabled'}.
        Values on the negative side of the normal are discarded if the plane is enabled.
    features : dict[str, array-like] or DataFrame
        Features table where each row corresponds to a label and each column
        is a feature. The first row corresponds to the background label.
    metadata : dict
        Layer metadata.
    multiscale : bool
        Whether the data is a multiscale image or not. Multiscale data is
        represented by a list of array like image data. If not specified by
        the user and if the data is a list of arrays that decrease in shape
        then it will be taken to be multiscale. The first image in the list
        should be the largest. Please note multiscale rendering is only
        supported in 2D. In 3D, only the lowest resolution scale is
        displayed.
    name : str
        Name of the layer.
    opacity : float
        Opacity of the layer visual, between 0.0 and 1.0.
    plane : dict or SlicingPlane
        Properties defining plane rendering in 3D. Properties are defined in
        data coordinates. Valid dictionary keys are
        {'position', 'normal', 'thickness', and 'enabled'}.
    properties : dict {str: array (N,)} or DataFrame
        Properties for each label. Each property should be an array of length
        N, where N is the number of labels, and the first property corresponds
        to background.
    rendering : str
        3D Rendering mode used by vispy. Must be one {'translucent', 'iso_categorical'}.
        'translucent' renders without lighting. 'iso_categorical' uses isosurface
        rendering to calculate lighting effects on labeled surfaces.
        The default value is 'iso_categorical'.
    rotate : float, 3-tuple of float, or n-D array.
        If a float convert into a 2D rotation matrix using that value as an
        angle. If 3-tuple convert into a 3D rotation matrix, using a yaw,
        pitch, roll convention. Otherwise assume an nD rotation. Angles are
        assumed to be in degrees. They can be converted from radians with
        np.degrees if needed.
    scale : tuple of float
        Scale factors for the layer.
    shear : 1-D array or n-D array
        Either a vector of upper triangular values, or an nD shear matrix with
        ones along the main diagonal.
    translate : tuple of float
        Translation values for the layer.
    visible : bool
        Whether the layer visual is currently being displayed.
        title : string, optional
        The title of the viewer window. By default 'napari'.
    ndisplay : {2, 3}, optional
        Number of displayed dimensions. By default 2.
    order : tuple of int, optional
        Order in which dimensions are displayed where the last two or last
        three dimensions correspond to row x column or plane x row x column if
        ndisplay is 2 or 3. By default None
    axis_labels : list of str, optional
        Dimension names. By default they are labeled with sequential numbers
    show : bool, optional
        Whether to show the viewer after instantiation. By default True.
    
    
    Returns
    -------
    viewer : :class:`napari.Viewer`
        The newly-created viewer.

Labels data#

The labels layer is a subclass of the Image layer and as such can support the same numpy-like arrays, including dask arrays, xarrays, and zarr arrays. A Labels layer though must be integer valued, and the background label must be 0.

Because the labels layer subclasses the image layer it inherits the great properties of the image layer, like supporting lazy loading and multiscale images for big data layers. For more information about both these concepts see the details in the image layer guide.

Creating a new labels layer#

As you can edit a Labels layer using the paintbrush and fill bucket, it is possible to create a brand-new empty labels layers by clicking the new labels layer button above the layers list. The shape of the new labels layer will match the size of any currently existing image layers, allowing you to paint on top of them.

Want to save without compression?

When saving a labels layer, lossless zlib compression is applied by default. To save with a different level of compression, consider using imageio.imwrite.
Adjusting compression can be accomplished by including the appropriate kwargs as outlined in the following locations for tiff or png files.

Non-editable mode#

If you want to disable editing of the labels layer you can set the editable property of the layer to False.

As note in the section on 3D rendering, when using 3D rendering the labels layer is not editable. Similarly, for now, a labels layer where the data is represented as a multiscale image is not editable.

3D rendering of labels#

All our layers can be rendered in both 2D and 3D mode, and one of our viewer buttons can toggle between each mode. The number of dimensions sliders will be 2 or 3 less than the total number of dimensions of the layer, allowing you to browse volumetric timeseries data and other high dimensional data.

import napari
from skimage import data
from scipy import ndimage as ndi

blobs = data.binary_blobs(length=128, volume_fraction=0.1, n_dim=3)
viewer = napari.view_image(blobs.astype(float), name='blobs')
labeled = ndi.label(blobs)[0]
labels_layer = viewer.add_labels(labeled, name='blob ID')
viewer.dims.ndisplay = 3

A 3D view of a labels layer on top of 3D blobs

Note though that when entering 3D rendering mode the colorpicker, paintbrush, and fill bucket options which allow for layer editing are all disabled. Those options are only supported when viewing a layer using 2D rendering.

Pan and zoom mode#

The default mode of the labels layer is to support panning and zooming, as in the image layer. This mode is represented by the magnifying glass in the layers control panel, and while it is selected editing the layer is not possible. Continue reading to learn how to use some of the editing modes. You can always return to pan and zoom mode by pressing the Z key when the labels layer is selected.

Shuffling label colors#

The color that each integer gets assigned is random, aside from 0 which always gets assigned to be transparent. The colormap we use is designed such that nearby integers get assigned distinct colors. The exact colors that get assigned as determined by a random seed and changing that seed will shuffle the colors that each label gets assigned. Changing the seed can be done by clicking on the shuffle colors button in the layers control panel. Shuffling colors can be useful as some colors may be hard to distinguish from the background or nearby objects.

Selecting a label#

A particular label can be selected either using the label combobox inside the layers control panel and typing in the value of the desired label or using the plus / minus buttons, or by selecting the color picker tool and then clicking on a pixel with the desired label in the image. When a label is selected you will see its integer inside the label combobox and the color or the label shown in the thumbnail next to the label combobox. If the 0 label is selected, then a checkerboard pattern is shown in the thumbnail to represent the transparent color.

You can quickly select the color picker by pressing the L key when the labels layer is selected.

You can set the selected label to be 0, the background label, by pressing E.

You can set the selected label to be one larger than the current largest label by pressing M. This selection will guarantee that you are then using a label that hasn’t been used before.

You can also increment or decrement the currently selected label by pressing the I or D key, respectively.

Painting in the labels layer#

One of the major use cases for the labels layer is to manually edit or create image segmentations. One of the tools that can be used for manual editing is the paintbrush, that can be made active from by clicking the paintbrush icon in the layers control panel. Once the paintbrush is enabled, the pan and zoom functionality of the viewer canvas gets disabled, and you can paint onto the canvas. You can temporarily re-enable pan and zoom by pressing and holding the spacebar. This feature can be useful if you want to move around the labels layer as you paint.

When you start painting the label that you are painting with, and the color that you will see are determined by the selected label. Note there is no explicit eraser tool, instead you just need to make the selected label 0 and then you are effectively painting with the background color. Remember you can use the color picker tool at any point to change the selected label.

You can adjust the size of your paintbrush using the brush size slider, making it as small as a single pixel for incredibly detailed painting.

If you have a multidimensional labels layer then your paintbrush will only edit data in the visible slice by default; however, if you enable the n_dimensional property and paintbrush then your paintbrush will extend out into neighbouring slices according to its size.

You can quickly select the paintbrush by pressing the P key when the labels layer is selected.

Using the fill bucket#

Sometimes you might want to replace an entire label with a different label. This could be because you want to make two touching regions have the same label, or you want to just replace one label with a different one, or maybe you have painted around the edge of a region and you want to quickly fill in its inside. To do this you can select the fill bucket by clicking on the droplet icon in the layer controls panel and then click on a target region of interest in the layer. The fill bucket will fill using the currently selected label.

By default, the fill bucket will only change contiguous or connected pixels of the same label as the pixel that is clicked on. If you want to change all the pixels of that label regardless of where they are in the slice, then you can set the contiguous property or checkbox to False.

If you have a multidimensional labels layer then your fill bucket will only edit data in the visible slice by default; however, if you enable the n_dimensional property and paintbrush then your fill bucket will extend out into neighbouring slices, either to all pixels with that label in the layer, or only connected pixels depending on if the contiguous property is disabled or not.

You can quickly select the fill bucket by pressing the F key when the labels layer is selected.

Creating, deleting, merging, and splitting connected components#

Using the color picker, paintbrush, and fill bucket tools one can create and edit object segmentation maps. Below we show how to use these tools to by perform common editing tasks on connected components (keep the contiguous box checked).

Drawing a connected component:

Press M to select a new label color. Select the paintbrush tool and draw a closed contour around the object. Select the fill bucket tool and click inside the contour to assign the label to all pixels of the object.

Selecting a connected component:

Select the background label with the color picker (alternative: press keyboard shortcut E), then use the fill bucket to set all pixels of the connected component to background.

Merging connected components:

Select the label of one of the components with the color picker tool and then filling the components to be merged with the fill bucket.

Splitting a connected component:

Splitting a connected component will introduce an additional object, therefore press M to select a label number that is not already in use. Use the paintbrush tool to draw a dividing line, then assign the new label to one of the parts with the fill bucket.

Undo / redo functionality#

When painting or using the fill bucket it can be easy to make a mistake that you might want to undo or then redo. For the labels layer we support an undo with ctrl-Z and redo with shift-ctrl-Z. We plan to support this sort of functionality more generally, but for now these actions will undo the most recent painting or filling event, up to 100 events in the past.

If you have multidimensional data, then adjusting the currently viewed slice will cause the undo history to be reset.