labelDefinitionCreatorMultiSensor

R2026b

Create label definitions for multi-sensor ground truth

Since R2026b

Description

The labelDefinitionCreatorMultiSensor object stores definitions of labels, sublabels, and attributes to label ground truth data for a multisignal workflow. Use Object Functions to add, remove, modify, or display label definitions. Use the create (Automated Driving Toolbox) object function to create a label definitions table from the labelDefinitionCreatorMultisignal object. You can use the created label definitions table with the groundTruthMultiSensor object or load the definitions into the Multi-Sensor Labeler app.

Creation

Syntax

ldc = labelDefinitionCreatorMultiSensor

ldc = labelDefinitionCreatorMultiSensor(labelDefs)

Description

ldc = labelDefinitionCreatorMultiSensor creates an empty label definition creator object ldc for a multi-sensor workflow. Add label definitions to this object using Object Functions. Use the info function to inspect details of stored labels, sublabels, and attributes.

ldc = labelDefinitionCreatorMultiSensor(labelDefs) creates a label definition creator object ldc for a multi-sensor workflow and stores definitions from the label definitions table labelDefs. Use Object Functions to add new label definitions or modify the existing label definitions. Use the info function to inspect details of stored labels, sublabels, and attributes.

Input Arguments

labelDefs — Label definitions table

table

Label definitions, specified as a table with up to nine columns. The possible columns are Name, SignalType, LabelType, Group, Description, LabelColor, PixelLabelID, SemanticPointLabelID, and Hierarchy. This table specifies the definitions of labels, sublabels, and attributes for labeling ground truth data. For more details, see the LabelDefinitions property of the groundTruthMultiSensor object.

Output Arguments

ldc — Label definition creator for multi-sensor workflow

labelDefinitionCreatorMultiSensor object

Label definition creator for the multi-sensor workflow, returned as a labelDefinitionCreatorMultiSensor object that contains information about label definitions associated with ground truth data.

Object Functions

addLabel	Add label to multi-sensor label definition creator
addSublabel	Add sublabel to label in multi-sensor label definition creator
addAttribute	Add attribute to label or sublabel in multi-sensor label definition creator

removeLabel	Remove label from multi-sensor label definition creator
removeSublabel	Remove sublabel from label in multi-sensor label definition creator
removeAttribute	Remove attribute from label or sublabel in multi-sensor label definition creator

editLabelGroup	Assign label to different group in multi-sensor label definition creator
editGroupName	Change group name in multi-sensor label definition creator
editLabelDescription	Modify label or sublabel description in multi-sensor label definition creator
editAttributeDescription	Modify attribute description in multi-sensor label definition creator

create

Create label definitions table from multi-sensor label definition creator

info	Display label, sublabel, or attribute information from multi-sensor label definition creator

Examples

Create Label Definition Creator and Add Label Definitions

Create a labelDefinitionCreatorMultiSensor object and add labels, sublabels, and attributes to it.

Create an empty labelDefinitionCreatorMultiSensor object.

ldc = labelDefinitionCreatorMultiSensor;

Add a label with the name "Vehicle". Specify the type as "Rectangle". Adding a "Rectangle" also adds a "Cuboid" entry to the label definitions table.

addLabel(ldc,"Vehicle","Rectangle")

Add an attribute with the name "Color" to the label "Vehicle". Specify the attribute type as a string with the value "Red".

addAttribute(ldc,"Vehicle","Color",attributeType.String,"Red")

Add a sublabel with the name "Wheel" to the label "Vehicle". Specify the type of the sublabel as "Rectangle".

addSublabel(ldc,"Vehicle","Wheel","Rectangle")

Add an attribute called "Diameter" to the sublabel "Wheel". Specify the attribute value as a "Numeric" scalar.

addAttribute(ldc,"Vehicle/Wheel","Diameter","Numeric",14)

Display the details of the updated labelDefinitionCreatorMultiSensor object.

ldc

ldc =
labelDefinitionCreatorMultiSensor contains the following labels:

	Vehicle with 1 sublabels and 1 attributes and belongs to None group.	(info)

For more details about attributes and sublabels, use the info method.

Create a label definitions table from the definitions stored in the object.

labelDefs = create(ldc)

labelDefs = 2×7 table
       Name        SignalType    LabelType     Group      Description    LabelColor     Hierarchy
    ___________    __________    _________    ________    ___________    __________    ____________

    {'Vehicle'}    Image         Rectangle    {'None'}       {' '}       {0×0 char}    {1×1 struct}
    {'Vehicle'}    PointCloud    Cuboid       {'None'}       {' '}       {0×0 char}    {1×1 struct}

Create Label Definition Creator from Existing Label Definitions Table

Create a labelDefinitionCreatorMultiSensor object from an existing label definitions table and modify it.

Load an existing multi-sensor label definitions table into the workspace.

labelDefFile = fullfile(toolboxdir("pointcloud"),"pcdata","labelDefsMultiSensor.mat");
ld = load(labelDefFile)

ld = 

  struct with fields:

    labelDefs: [12×8 table]

Create a labelDefinitionCreatorMultiSensor object from the label definitions table.

ldc = labelDefinitionCreatorMultiSensor(ld.labelDefs)

labelDefinitionCreatorMultiSensor contains the following labels:

	Car with 0 sublabels and 0 attributes and belongs to None group.	(info)
	Face with 0 sublabels and 0 attributes and belongs to None group.	(info)
	Lane with 0 sublabels and 0 attributes and belongs to None group.	(info)
	Building with 0 sublabels and 0 attributes and belongs to None group.	(info)
	Sky with 0 sublabels and 0 attributes and belongs to None group.	(info)
	Road with 0 sublabels and 0 attributes and belongs to None group.	(info)
	Truck with 0 sublabels and 0 attributes and belongs to None group.	(info)
	RedLight with 0 sublabels and 0 attributes and belongs to TrafficLight group.	(info)
	GreenLight with 0 sublabels and 0 attributes and belongs to TrafficLight group.	(info)
	YellowLight with 0 sublabels and 0 attributes and belongs to TrafficLight group.	(info)

For more details about attributes and sublabels, use the info method.

Add a new attribute to the label "Car".

addAttribute(ldc,"Car","Color","List",{"Red","Green","Blue"})

Display the details of the updated labelDefinitionCreatorMultiSensor object.

ldc

ldc = 

labelDefinitionCreatorMultiSensor contains the following labels:

	Car with 0 sublabels and 1 attributes and belongs to None group.	(info)
	Face with 0 sublabels and 0 attributes and belongs to None group.	(info)
	Lane with 0 sublabels and 0 attributes and belongs to None group.	(info)
	Building with 0 sublabels and 0 attributes and belongs to None group.	(info)
	Sky with 0 sublabels and 0 attributes and belongs to None group.	(info)
	Road with 0 sublabels and 0 attributes and belongs to None group.	(info)
	Truck with 0 sublabels and 0 attributes and belongs to None group.	(info)
	RedLight with 0 sublabels and 0 attributes and belongs to TrafficLight group.	(info)
	GreenLight with 0 sublabels and 0 attributes and belongs to TrafficLight group.	(info)
	YellowLight with 0 sublabels and 0 attributes and belongs to TrafficLight group.	(info)

For more details about attributes and sublabels, use the info method.

More About

Supported Label Types

The labelDefinitionCreatorMultiSensor object supports these label types when creating label definitions:

Label Type	Signal Type	Description
labelType.Rectangle	Image	Axis-aligned rectangular ROI for image signals
labelType.RotatedRectangle	Image	Rotated rectangular ROI for image signals
labelType.Line	Image PointCloud	Polyline ROI for image and point cloud signals
labelType.Polygon	Image	Polygon ROI for image signals
labelType.PixelLabel	Image	Pixel-level semantic label for image signals
labelType.ProjectedCuboid	Image	3-D cuboid projected onto image signals
labelType.Point	Image	Point ROI for image signals
labelType.Cuboid	PointCloud	3-D cuboid ROI for point cloud signals
labelType.SemanticPointLabel	PointCloud	Per-point semantic label for point cloud signals. Use this type for 3-D semantic segmentation of point clouds.
labelType.Scene	Time	Scene-level label that applies across all signals for a time range

Version History

Introduced in R2026b