fiftyone.core.plots.matplotlib#

Matplotlib plots.

Copyright 2017-2025, Voxel51, Inc.

Functions:

plot_confusion_matrix(confusion_matrix,Β labels)

Plots a confusion matrix.

plot_regressions(ytrue,Β ypred[,Β samples,Β ...])

Plots the given regression results.

plot_pr_curve(precision,Β recall[,Β label,Β ...])

Plots a precision-recall (PR) curve.

plot_pr_curves(precisions,Β recall,Β classes)

Plots a set of per-class precision-recall (PR) curves.

plot_roc_curve(fpr,Β tpr[,Β roc_auc,Β title,Β ...])

Plots a receiver operating characteristic (ROC) curve.

lines([x,Β y,Β samples,Β ids,Β link_field,Β ...])

Plots the given lines(s) data.

scatterplot(points[,Β samples,Β ids,Β ...])

Generates an interactive scatterplot of the given points.

location_scatterplot([locations,Β samples,Β ...])

Generates an interactive scatterplot of the given location coordinates with a map rendered in the background of the plot.

Classes:

InteractiveMatplotlibPlot(figure,Β **kwargs)

Base class for interactive matplotlib plots.

InteractiveCollection(collection[,Β ids,Β ...])

Interactive wrapper for a matplotlib collection.

fiftyone.core.plots.matplotlib.plot_confusion_matrix(confusion_matrix, labels, show_values=True, show_colorbar=True, cmap=None, title=None, xticks_rotation=45.0, values_format=None, ax=None, figsize=None)#

Plots a confusion matrix.

Parameters:
  • confusion_matrix – a num_true x num_preds confusion matrix

  • labels – a max(num_true, num_preds) array-like of class labels

  • show_values (True) – whether to show counts in the confusion matrix cells

  • show_colorbar (True) – whether to show a colorbar

  • cmap (None) – a colormap recognized by matplotlib

  • title (None) – a title for the plot

  • xticks_rotation (45.0) – a rotation for the x-tick labels. Can be numeric degrees, β€œvertical”, β€œhorizontal”, or None

  • values_format (None) – a format string like ".2g" or "d" to use to format the cell counts

  • ax (None) – a matplotlib axis to plot in

  • figsize (None) – a (width, height) for the figure, in inches

Returns:

a matplotlib figure

fiftyone.core.plots.matplotlib.plot_regressions(ytrue, ypred, samples=None, ids=None, labels=None, sizes=None, classes=None, gt_field=None, pred_field=None, best_fit_label=None, marker_size=None, cmap=None, title=None, ax=None, figsize=None, style=None, **kwargs)#

Plots the given regression results.

Parameters:
  • ytrue – an array-like of ground truth values

  • ypred – an array-like of predicted values

  • samples (None) – the fiftyone.core.collections.SampleCollection whose data is being visualized

  • ids (None) – an array-like of sample or frame IDs corresponding to the regressions. If not provided but samples are provided, the appropriate IDs will be extracted from the samples

  • labels (None) –

    data to use to color the points. Can be any of the following:

  • sizes (None) –

    data to use to scale the sizes of the points. Can be any of the following:

  • classes (None) – a list of classes whose points to plot. Only applicable when labels contains strings

  • gt_field (None) – the name of the ground truth field

  • pred_field (None) – the name of the predictions field

  • best_fit_label (None) – a custom legend label for the best fit line

  • marker_size (None) – the marker size to use. If sizes are provided, this value is used as a reference to scale the sizes of all points

  • cmap (None) – a colormap recognized by matplotlib

  • title (None) – a title for the plot

  • ax (None) – a matplotlib axis to plot in

  • figsize (None) – a (width, height) for the figure, in inches

  • style (None) – a style to use for the plot

  • **kwargs – optional keyword arguments for matplotlib’s scatter()

Returns:

a matplotlib figure

fiftyone.core.plots.matplotlib.plot_pr_curve(precision, recall, label=None, title=None, ax=None, figsize=None, style=None, **kwargs)#

Plots a precision-recall (PR) curve.

Parameters:
  • precision – an array-like of precision values

  • recall – an array-like of recall values

  • label (None) – a label for the curve

  • title (None) – a title for the plot

  • ax (None) – a matplotlib axis to plot in

  • figsize (None) – a (width, height) for the figure, in inches

  • style (None) – a style to use for the plot

  • **kwargs – optional keyword arguments for matplotlib’s plot()

Returns:

a matplotlib figure

fiftyone.core.plots.matplotlib.plot_pr_curves(precisions, recall, classes, title=None, ax=None, figsize=None, style=None, **kwargs)#

Plots a set of per-class precision-recall (PR) curves.

Parameters:
  • precisions – a num_classes x num_recalls array-like of per-class precision values

  • recall – an array-like of recall values

  • classes – the list of classes

  • title (None) – a title for the plot

  • ax (None) – a matplotlib axis to plot in

  • figsize (None) – a (width, height) for the figure, in inches

  • style (None) – a style to use for the plot

  • **kwargs – optional keyword arguments for matplotlib’s plot()

Returns:

a matplotlib figure

fiftyone.core.plots.matplotlib.plot_roc_curve(fpr, tpr, roc_auc=None, title=None, ax=None, figsize=None, style=None, **kwargs)#

Plots a receiver operating characteristic (ROC) curve.

Parameters:
  • fpr – an array-like of false positive rates

  • tpr – an array-like of true positive rates

  • roc_auc (None) – the area under the ROC curve

  • title (None) – a title for the plot

  • ax (None) – a matplotlib axis to plot in

  • figsize (None) – a (width, height) for the figure, in inches

  • style (None) – a style to use for the plot

  • **kwargs – optional keyword arguments for matplotlib’s plot()

Returns:

a matplotlib figure

fiftyone.core.plots.matplotlib.lines(x=None, y=None, samples=None, ids=None, link_field=None, sizes=None, labels=None, colors=None, marker_size=None, title=None, xlabel=None, ylabel=None, ax=None, ax_equal=False, figsize=None, style=None, buttons=None, **kwargs)#

Plots the given lines(s) data.

You can attach plots generated by this method to an App session via its fiftyone.core.session.Session.plots attribute, which will automatically sync the session’s view with the currently selected points in the plot. To enable this functionality, you must pass samples to this method.

You can use the sizes parameter to scale the sizes of the points.

Parameters:
  • x (None) –

    the x data to plot. Can be any of the following:

    • an array-like of values

    • a num_lines x n array-like or list of length num_lines of array-likes of values for multiple line traces

    • the name of a sample field or embedded.field.name of samples from which to extract values for a single line

    • the name of a frame field or frames.embedded.field.name of samples from which to extract values for per-sample line traces

    • a fiftyone.core.expressions.ViewExpression that resolves to a list (one line plot) or list of lists (multiple line plots) of numeric values to compute from samples via fiftyone.core.collections.SampleCollection.values()

  • y (None) –

    the y data to plot. Can be any of the following:

    • an array-like of values

    • a num_lines x n array-like or list of length num_lines of array-likes of values for multiple line traces

    • the name of a sample field or embedded.field.name of samples from which to extract values for a single line

    • the name of a frame field or frames.embedded.field.name of samples from which to extract values for per-sample line traces

    • a fiftyone.core.expressions.ViewExpression that resolves to a list (one line plot) or list of lists (multiple line plots) of numeric values to compute from samples via fiftyone.core.collections.SampleCollection.values()

  • samples (None) – the fiftyone.core.collections.SampleCollection whose data is being visualized

  • ids (None) – an array-like of IDs of same shape as y. If not provided but samples are provided, the appropriate IDs will be extracted from the samples

  • link_field (None) –

    a field of samples whose data corresponds to y. Can be any of the following:

    • None, if the line data correspond to samples (single trace) or frames (multiple traces)

    • "frames", if the line data correspond to frames (multiple traces). This option exists only for consistency with other plotting methods; in practice, it will be automatically inferred whenever multiple traces are being plotted

    • the name of a fiftyone.core.labels.Label field, if the line data correspond to the labels in this field

  • sizes (None) –

    data to use to scale the sizes of the points. Can be any of the following:

  • labels (None) – a label or list of labels for the line traces

  • colors (None) – a list of colors recognized by matplotlib to use for the line traces. See https://matplotlib.org/stable/tutorials/colors/colormaps.html for more information

  • marker_size (None) – the marker size to use. If sizes are provided, this value is used as a reference to scale the sizes of all points

  • title (None) – a title for the plot

  • xlabel (None) – an x-axis label

  • ylabel (None) – a y-axis label

  • ax (None) – a matplotlib axis to plot in

  • ax_equal (False) – whether to set axis("equal")

  • figsize (None) – a (width, height) for the figure, in inches

  • style (None) – a style to use for the plot

  • buttons (None) – a list of (label, icon_image, callback) tuples defining buttons to add to the plot

  • **kwargs – optional keyword arguments for matplotlib’s plot() and scatter()

Returns:

one of the following

fiftyone.core.plots.matplotlib.scatterplot(points, samples=None, ids=None, link_field=None, labels=None, sizes=None, classes=None, marker_size=None, cmap=None, title=None, ax=None, ax_equal=False, figsize=None, style=None, buttons=None, **kwargs)#

Generates an interactive scatterplot of the given points.

You can attach plots generated by this method to an App session via its fiftyone.core.session.Session.plots attribute, which will automatically sync the session’s view with the currently selected points in the plot. To enable this functionality, you must pass samples to this method.

This method supports 2D or 3D visualizations, but interactive point selection is only available in 2D.

You can use the labels parameters to define a coloring for the points, and you can use the sizes parameter to scale the sizes of the points.

Parameters:
  • points – a num_points x num_dims array-like of points

  • samples (None) – the fiftyone.core.collections.SampleCollection whose data is being visualized

  • ids (None) – an array-like of IDs corresponding to the points. If not provided but samples are provided, the appropriate IDs will be extracted from the samples

  • link_field (None) –

    a field of samples whose data corresponds to points. Can be any of the following:

    • None, if the points correspond to samples

    • "frames", if the points correspond to frames

    • the name of a fiftyone.core.labels.Label field, if the points correspond to the labels in this field

  • labels (None) –

    data to use to color the points. Can be any of the following:

  • sizes (None) –

    data to use to scale the sizes of the points. Can be any of the following:

  • classes (None) – a list of classes whose points to plot. Only applicable when labels contains strings

  • marker_size (None) – the marker size to use. If sizes are provided, this value is used as a reference to scale the sizes of all points

  • cmap (None) – a colormap recognized by matplotlib

  • title (None) – a title for the plot

  • ax (None) – a matplotlib axis to plot in

  • ax_equal (False) – whether to set axis("equal")

  • figsize (None) – a (width, height) for the figure, in inches

  • style (None) – a style to use for the plot

  • buttons (None) – a list of (label, icon_image, callback) tuples defining buttons to add to the plot

  • **kwargs – optional keyword arguments for matplotlib’s scatter()

Returns:

one of the following

fiftyone.core.plots.matplotlib.location_scatterplot(locations=None, samples=None, ids=None, labels=None, sizes=None, classes=None, map_type='satellite', show_scale_bar=False, api_key=None, marker_size=None, cmap=None, title=None, ax=None, ax_equal=False, figsize=None, style=None, buttons=None, **kwargs)#

Generates an interactive scatterplot of the given location coordinates with a map rendered in the background of the plot.

Location data is specified via the locations parameter.

You can attach plots generated by this method to an App session via its fiftyone.core.session.Session.plots attribute, which will automatically sync the session’s view with the currently selected points in the plot. To enable this functionality, you must pass samples to this method.

You can use the labels parameters to define a coloring for the points, and you can use the sizes parameter to scale the sizes of the points.

Parameters:
  • locations (None) –

    the location data to plot. Can be any of the following:

    • None, in which case samples must have a single fiftyone.core.labels.GeoLocation field whose point attribute contains location data

    • a num_locations x 2 array-like of (longitude, latitude) coordinates

    • the name of a fiftyone.core.labels.GeoLocation field of samples with (longitude, latitude) coordinates in its point attribute

  • samples (None) – the fiftyone.core.collections.SampleCollection whose data is being visualized

  • ids (None) – an array-like of IDs corresponding to the locations. If not provided but samples are provided, the appropriate IDs will be extracted from the samples

  • labels (None) –

    data to use to color the points. Can be any of the following:

  • sizes (None) –

    data to use to scale the sizes of the points. Can be any of the following:

  • classes (None) – a list of classes whose points to plot. Only applicable when labels contains strings

  • map_type ("satellite") – the map type to render. Supported values are ("roadmap", "satellite", "hybrid", "terrain")

  • show_scale_bar (False) – whether to render a scale bar on the plot

  • api_key (None) – a Google Maps API key to use

  • marker_size (None) – the marker size to use. If sizes are provided, this value is used as a reference to scale the sizes of all points

  • cmap (None) – a colormap recognized by matplotlib

  • title (None) – a title for the plot

  • ax (None) – a matplotlib axis to plot in

  • ax_equal (False) – whether to set axis("equal")

  • figsize (None) – a (width, height) for the figure, in inches

  • style (None) – a style to use for the plot

  • buttons (None) – a list of (label, icon_image, callback) tuples defining buttons to add to the plot

  • **kwargs – optional keyword arguments for matplotlib’s scatter()

Returns:

one of the following

class fiftyone.core.plots.matplotlib.InteractiveMatplotlibPlot(figure, **kwargs)#

Bases: InteractivePlot

Base class for interactive matplotlib plots.

Parameters:

Attributes:

supports_session_updates

Whether this plot supports automatic updates in response to session changes.

init_view

A fiftyone.core.collections.SampleCollection defining the initial view from which to derive selection views when points are selected in the plot.

is_connected

Whether this plot is currently connected.

is_disconnected

Whether this plot is currently disconnected.

is_frozen

Whether this plot is currently frozen.

link_type

The link type between this plot and a connected session.

selected_ids

A list of IDs of the currently selected points.

selection_mode

The current selection mode of the plot.

Methods:

show()

Shows this plot.

save(path[,Β dpi])

Saves the plot as an image.

connect()

Connects this plot, if necessary.

disconnect()

Disconnects the plot, if necessary.

freeze()

Freezes the plot, replacing it with a static image.

recommend_link_type([label_field,Β samples])

Recommends a link type for the given info.

register_disconnect_callback(callback)

Registers a callback that can disconnect this plot from a SessionPlot connected to it.

register_selection_callback(callback)

Registers a selection callback for this plot.

register_sync_callback(callback)

Registers a callback that can sync this plot with a SessionPlot connected to it.

reset()

Resets the plot to its default state.

select_ids(ids[,Β view])

Selects the points with the given IDs in this plot.

property supports_session_updates#

Whether this plot supports automatic updates in response to session changes.

show()#

Shows this plot.

save(path, dpi=None, **kwargs)#

Saves the plot as an image.

Parameters:
  • path – the path to write the image

  • dpi (None) – a resolution in dots per inch

  • **kwargs – keyword arguments for matplotlib.pyplot.savefig

connect()#

Connects this plot, if necessary.

disconnect()#

Disconnects the plot, if necessary.

freeze()#

Freezes the plot, replacing it with a static image.

The plot will also be disconnected.

Only applicable in notebook contexts.

property init_view#

A fiftyone.core.collections.SampleCollection defining the initial view from which to derive selection views when points are selected in the plot.

This view will also be shown when the plot is in its default state (no selection).

property is_connected#

Whether this plot is currently connected.

property is_disconnected#

Whether this plot is currently disconnected.

property is_frozen#

Whether this plot is currently frozen.

The link type between this plot and a connected session.

Recommends a link type for the given info.

Parameters:
Returns:

a (link_type, label_fields, selection_mode, init_fcn) tuple

register_disconnect_callback(callback)#

Registers a callback that can disconnect this plot from a SessionPlot connected to it.

The typical use case for this function is to serve as the callback for a disconnect button on the plot.

Parameters:

callback – a function with no arguments

register_selection_callback(callback)#

Registers a selection callback for this plot.

Selection callbacks are functions that take a single argument containing the list of currently selected IDs.

If a selection callback is registered, this plot should invoke it each time their selection is updated.

Parameters:

callback – a selection callback

register_sync_callback(callback)#

Registers a callback that can sync this plot with a SessionPlot connected to it.

The typical use case for this function is to serve as the callback for a sync button on the plot.

Parameters:

callback – a function with no arguments

reset()#

Resets the plot to its default state.

select_ids(ids, view=None)#

Selects the points with the given IDs in this plot.

Parameters:
  • ids – a list of IDs, or None to reset the plot to its default state

  • view (None) – the fiftyone.core.view.DatasetView corresponding to the given IDs, if available

property selected_ids#

A list of IDs of the currently selected points.

An empty list means all points are deselected, and None means default state (nothing selected or unselected).

If the plot is not connected, returns None.

property selection_mode#

The current selection mode of the plot.

This property controls how the current view is updated in response to updates from InteractivePlot instances that are linked to labels or frames.

When link_type is "frames", the supported values are:

  • "select": show video samples with labels only for the selected frames

  • "match": show unfiltered video samples containing at least one selected frame

  • "frames": show only the selected frames in a frames view

When link_type is "labels", the supported values are:

  • "select": show only the selected labels

  • "match": show unfiltered samples containing at least one selected label

  • "patches": show the selected labels in a patches view

Note

In order to use "patches" selection mode, you must have provided an init_fcn when constructing this plot that defines how to create the base patches view. This usually involves to_patches() or to_evaluation_patches()

Note

In order to use "frames" selection mode, you must have provided an init_fcn when constructing this plot that defines how to create the base frames view. This usually involves to_frames()

class fiftyone.core.plots.matplotlib.InteractiveCollection(collection, ids=None, buttons=None, alpha_other=0.25, expand_selected=3.0, click_tolerance=0.02, **kwargs)#

Bases: InteractiveMatplotlibPlot

Interactive wrapper for a matplotlib collection.

This class enables collection points to be lasso-ed and click selected.

The currently selected points are given a visually distinctive style, and you can modify your selection by either clicking on individual points or drawing a lasso around new points.

When the shift key is pressed, new selections are added to the selected set, or subtracted if the new selection is a subset of the current selection.

Parameters:
  • collection – a matplotlib.collections.Collection to select points from

  • ids (None) – an array-like of IDs corresponding to the points in collection

  • buttons (None) – a list of (label, icon_image, callback) tuples defining buttons to add to the plot

  • alpha_other (0.25) – a transparency value for unselected points

  • expand_selected (3.0) – expand the size of selected points by this multiple

  • click_tolerance (0.02) – a click distance tolerance in [0, 1] when clicking individual points

  • **kwargs – keyword arguments for the fiftyone.core.plots.base.InteractivePlot constructor

Methods:

connect()

Connects this plot, if necessary.

disconnect()

Disconnects the plot, if necessary.

freeze()

Freezes the plot, replacing it with a static image.

recommend_link_type([label_field,Β samples])

Recommends a link type for the given info.

register_disconnect_callback(callback)

Registers a callback that can disconnect this plot from a SessionPlot connected to it.

register_selection_callback(callback)

Registers a selection callback for this plot.

register_sync_callback(callback)

Registers a callback that can sync this plot with a SessionPlot connected to it.

reset()

Resets the plot to its default state.

save(path[,Β dpi])

Saves the plot as an image.

select_ids(ids[,Β view])

Selects the points with the given IDs in this plot.

show()

Shows this plot.

Attributes:

init_view

A fiftyone.core.collections.SampleCollection defining the initial view from which to derive selection views when points are selected in the plot.

is_connected

Whether this plot is currently connected.

is_disconnected

Whether this plot is currently disconnected.

is_frozen

Whether this plot is currently frozen.

link_type

The link type between this plot and a connected session.

selected_ids

A list of IDs of the currently selected points.

selection_mode

The current selection mode of the plot.

supports_session_updates

Whether this plot supports automatic updates in response to session changes.

connect()#

Connects this plot, if necessary.

disconnect()#

Disconnects the plot, if necessary.

freeze()#

Freezes the plot, replacing it with a static image.

The plot will also be disconnected.

Only applicable in notebook contexts.

property init_view#

A fiftyone.core.collections.SampleCollection defining the initial view from which to derive selection views when points are selected in the plot.

This view will also be shown when the plot is in its default state (no selection).

property is_connected#

Whether this plot is currently connected.

property is_disconnected#

Whether this plot is currently disconnected.

property is_frozen#

Whether this plot is currently frozen.

The link type between this plot and a connected session.

Recommends a link type for the given info.

Parameters:
Returns:

a (link_type, label_fields, selection_mode, init_fcn) tuple

register_disconnect_callback(callback)#

Registers a callback that can disconnect this plot from a SessionPlot connected to it.

The typical use case for this function is to serve as the callback for a disconnect button on the plot.

Parameters:

callback – a function with no arguments

register_selection_callback(callback)#

Registers a selection callback for this plot.

Selection callbacks are functions that take a single argument containing the list of currently selected IDs.

If a selection callback is registered, this plot should invoke it each time their selection is updated.

Parameters:

callback – a selection callback

register_sync_callback(callback)#

Registers a callback that can sync this plot with a SessionPlot connected to it.

The typical use case for this function is to serve as the callback for a sync button on the plot.

Parameters:

callback – a function with no arguments

reset()#

Resets the plot to its default state.

save(path, dpi=None, **kwargs)#

Saves the plot as an image.

Parameters:
  • path – the path to write the image

  • dpi (None) – a resolution in dots per inch

  • **kwargs – keyword arguments for matplotlib.pyplot.savefig

select_ids(ids, view=None)#

Selects the points with the given IDs in this plot.

Parameters:
  • ids – a list of IDs, or None to reset the plot to its default state

  • view (None) – the fiftyone.core.view.DatasetView corresponding to the given IDs, if available

property selected_ids#

A list of IDs of the currently selected points.

An empty list means all points are deselected, and None means default state (nothing selected or unselected).

If the plot is not connected, returns None.

property selection_mode#

The current selection mode of the plot.

This property controls how the current view is updated in response to updates from InteractivePlot instances that are linked to labels or frames.

When link_type is "frames", the supported values are:

  • "select": show video samples with labels only for the selected frames

  • "match": show unfiltered video samples containing at least one selected frame

  • "frames": show only the selected frames in a frames view

When link_type is "labels", the supported values are:

  • "select": show only the selected labels

  • "match": show unfiltered samples containing at least one selected label

  • "patches": show the selected labels in a patches view

Note

In order to use "patches" selection mode, you must have provided an init_fcn when constructing this plot that defines how to create the base patches view. This usually involves to_patches() or to_evaluation_patches()

Note

In order to use "frames" selection mode, you must have provided an init_fcn when constructing this plot that defines how to create the base frames view. This usually involves to_frames()

show()#

Shows this plot.

property supports_session_updates#

Whether this plot supports automatic updates in response to session changes.