Ground Truth Labeling

Interactive ground truth labeling of multiple signals

Use the Ground Truth Labeler app to label multiple signals representing the same scene. You can label videos, image sequences, and lidar signals such as point cloud sequences. Use labeled ground truth as training data for machine learning and deep learning models, such as object detectors or semantic segmentation networks. To automate the labeling of ground truth data, you can use a built-in automation algorithm or develop your own algorithm. The app also provides APIs for displaying additional time-synchronized signals and for loading custom data sources. To get started labeling a scene, see Get Started with Ground Truth Labelling.

Apps

Ground Truth Labeler

Label ground truth data for automated driving applications

Functions

expand all

Create Labels and Label Definitions

Load Data

`groundTruthMultisignal`	Ground truth label data for multiple signals
`ROILabelData`	Ground truth data for ROI labels
`SceneLabelData`	Ground truth data for scene labels
`vision.labeler.loading.MultiSignalSource`	Interface for loading signal data into Ground Truth Labeler app
`vision.labeler.loading.VideoSource`	Load signals from video sources into Ground Truth Labeler app
`vision.labeler.loading.ImageSequenceSource`	Load signals from image sequence sources into Ground Truth Labeler app
`vision.labeler.loading.VelodyneLidarSource`	Load signals from Velodyne lidar sources into Ground Truth Labeler app
`vision.labeler.loading.RosbagSource`	Load signals from rosbag sources into Ground Truth Labeler app
`vision.labeler.loading.PointCloudSequenceSource`	Load signals from point cloud sequence sources into Ground Truth Labeler app
`vision.labeler.loading.CustomImageSource`	Load signals from custom image sources into Ground Truth Labeler app
`vision.labeler.loading.SignalType`	Signal type enumerations for labeling

Create Label Definitions

`labelDefinitionCreatorMultisignal`	Object for storing, modifying, and creating label definitions table for multisignal workflow
`create`	Create label definitions table from label definition creator object for multisignal workflow
`info`	Display label, sublabel, or attribute information stored in label definition creator object for multisignal workflow

Add

`addLabel`	Add label to label definition creator object for multisignal workflow
`addSublabel`	Add sublabel to label in label definition creator object for multisignal workflow
`addAttribute`	Add attributes to label or sublabel in label definition creator object for multisignal workflow

Edit

`editGroupName`	Change group name in label definition creator object for multisignal workflow
`editLabelGroup`	Modify label group name in label definition creator object for multisignal workflow
`editLabelDescription`	Modify label or sublabel description in label definition creator object for multisignal workflow
`editAttributeDescription`	Modify attribute description in label definition creator object for multisignal workflow

Remove

`removeLabel`	Remove label from label definition creator object for multisignal workflow
`removeSublabel`	Remove sublabel from label in label definition creator object for multisignal workflow
`removeAttribute`	Remove attribute from label or sublabel in label definition creator object for multisignal workflow

Automate Labeling

Action Flags

`initialize`	Initialize state for algorithm execution (optional)
`run`	Run label automation on every frame in interval
`terminate`	Terminate automated algorithm (optional)

Validation

`checkSetup`	Set up validation (optional)
`checkSignalType`	Validate signal type
`checkLabelDefinition`	Validate label definition
`supportsMultisignalAutomation`	Set multisignal algorithm automation flag

Algorithm

`vision.labeler.AutomationAlgorithm`	Interface for algorithm automation in ground truth labeling
`settingsDialog`	Display algorithm settings (optional)
`vision.labeler.mixin.Temporal`	Mixin interface for adding temporal context to automation algorithms
`supportsReverseAutomation`	Set reverse algorithm automation flag

Export and Postprocess Labels

Store Data

`groundTruthMultisignal`	Ground truth label data for multiple signals
`ROILabelData`	Ground truth data for ROI labels
`SceneLabelData`	Ground truth data for scene labels
`changeFilePaths`	Change file paths in multisignal ground truth data

Select Labels

`selectLabelsByLabelName`	Select multisignal ground truth by label name
`selectLabelsByLabelType`	Select multisignal ground truth by label type
`selectLabelsByGroupName`	Select multisignal ground truth by label group name
`selectLabelsBySignalName`	Select multisignal ground truth by signal name
`selectLabelsBySignalType`	Select multisignal ground truth labels by signal type

Enumerate Attribute and Label Types

`labelType`	Label type enumerations for labeling
`attributeType`	Attribute type enumerations for labeling

Create Training Data

`gatherLabelData`	Gather synchronized label data from ground truth
`writeFrames`	Write signal frames for ground truth data to disk

Connect External Display Tool

`driving.connector.Connector`	Interface to connect external tool to Ground Truth Labeler app
`frameChangeListener`	Update external tool when new frame is displayed in Ground Truth Labeler app
`dataSourceChangeListener`	Update external tool when connecting to signal being loaded into Ground Truth Labeler app
`labelLoadListener`	Update external tool for new label data in Ground Truth Labeler app
`labelDefinitionLoadListener`	Update external tool for new label definitions in Ground Truth Labeler app
`close`	Close external tool connected to Ground Truth Labeler app
`addLabelData`	Add custom label data at current time
`queryLabelData`	Query for custom label data at current time
`updateLabelerCurrentTime`	Update current time in Ground Truth Labeler app
`disconnect`	Disconnect external tool from Ground Truth Labeler app

Topics

Get Started

Choose an App to Label Ground Truth Data
Decide which app to use to label ground truth data: Image Labeler, Video Labeler, Ground Truth Labeler, Lidar Labeler, Signal Labeler, or Medical Image Labeler.
Get Started with Ground Truth Labelling

Interactively label multiple lidar and video signals simultaneously.
- STEP 1: Load Ground Truth Signals to Label
- STEP 2: Label Ground Truth for Multiple Signals
- STEP 3: Export and Explore Ground Truth Labels for Multiple Signals
Keyboard Shortcuts and Mouse Actions for Ground Truth Labeler
Use keyboard shortcuts and mouse actions to increase productivity while using the Ground Truth Labeler app.

Load Ground Truth Data Sources

Sources vs. Signals in Ground Truth Labeling
In the Ground Truth Labeler app, a data source is a file or folder containing one or more signals to load.
Create Class for Loading Custom Ground Truth Data Sources
Learn how to create a class for loading signals from custom data sources into the Ground Truth Labeler app.
Use Custom Image Source Reader for Labeling
Create a reader function to load a custom image data source into the Ground Truth Labeler or Video Labeler.
Control Playback of Signal Frames for Labeling
Control which signal frames display in the Ground Truth Labeler app during playback.

Label Ground Truth

Label Pixels for Semantic Segmentation
Label pixels for training a semantic segmentation network by using a labeling app.
Label Lidar Point Clouds for Object Detection
Label lidar point clouds to use as training data for object detectors.
Use Sublabels and Attributes to Label Ground Truth Data
Learn how ROI sublabels and attributes work in labeling apps.
View Summary of Ground Truth Labels
View a summary of ROI and scene labels in a labeling app session.

Automate Labeling

Create Automation Algorithm for Labeling
Create a custom automation algorithm to use in a labeling app.
Temporal Automation Algorithms
Create a time-based custom tracking algorithm to import into a labeling app.

Use Ground Truth Labels

Share and Store Labeled Ground Truth Data
Share and store labeled ground truth data exported from labeling apps.
How Labeler Apps Store Exported Pixel Labels
Learn how the labeling apps store pixel label data.

Featured Examples

Automate Ground Truth Labeling Across Multiple Signals

Automate the labeling of image and point cloud signals simultaneously by using a vehicle detector and camera-to-lidar calibration.

Open Live Script

Automate Ground Truth Labeling of Lane Boundaries

Develop an algorithm for the automated marking of lane boundaries in the Ground Truth Labeler app.

Open Live Script

Automate Ground Truth Labeling for Semantic Segmentation

Use a pretrained semantic segmentation algorithm to segment an image, and use this algorithm to automate ground truth labeling.

Open Live Script

Automate Attributes of Labeled Objects

Use a vehicle detection and distance estimation algorithm to automate the labeling of objects and their attributes.

Open Live Script

Evaluate Lane Boundary Detections Against Ground Truth Data

Compare ground truth data against the results of a lane boundary detection algorithm.

Open Live Script

Evaluate and Visualize Lane Boundary Detections Against Ground Truth

Evaluate the performance of lane boundary detection against known ground truth using visualization tools.

Open Script