Create Custom Automation Algorithm for Labeling

This topic shows how to implement a custom automation algorithm using a class-based interface to accelerate ground truth labeling across labeling apps such as Image Labeler, Video Labeler, Lidar Labeler (Lidar Toolbox), and Ground Truth Labeler (Automated Driving Toolbox). The labeling apps enable you to label ground truth for a variety of data sources and automatically label your data by creating and importing a custom automation algorithm.

To build a custom automation algorithm for labeling apps, you can use either a function-based or class-based interface. The apps provide templates for both interface types. The table below compares their capabilities to help you choose the right interface for your workflow:

Automation Algorithm Interface Type	Description	Use Case & Capabilities
Function-based	Define automation logic using a standalone function with parameter tuning.	Ideal for quick setup and prototyping simple automation tasks Useful for migrating existing functional code Lightweight and fast to implement
Class-based	Create a custom class inheriting from `vision.labeler.AutomationAlgorithm`, which offers full control over labeling behavior and app interaction.	Suitable for complex workflows requiring customization Provides access to temporal information using `vision.labeler.mixin.Temporal` class. Supports automating `blockedImage` labels using `vision.labeler.mixin.BlockedImageAutomation` class. Provides customized initialization and termination steps. Enables implementation of customized settings dialog. Enables customized algorithm state management across image frames. Supports multiple signals or point cloud signals, such as what you might use with the Ground Truth Labeler (Automated Driving Toolbox). Customized name, description algorithm instructions.

For more information about the function-based automation algorithm interface, see Create Automation Algorithm Function for Labeling.

Create and Implement Automation Algorithm

The vision.labeler.AutomationAlgorithm class enables you to define a custom label automation algorithm for use in the labeling apps. You can use the class to define the interface used by the app to run an automation algorithm.

To define and use a custom automation algorithm, you must first define a class for your algorithm and save it to the appropriate folder.

Create Automation Class Template from Labeler App

At the MATLAB^® command prompt, enter one of these commands to open the relevant labeling app.
```
imageLabeler
videoLabeler
lidarLabeler
groundTruthLabeler
```
Load a data source and create at least one label definition.
On the app toolstrip, select Select Algorithm > Add Algorithm > Create New Algorithm.
This opens the vision.labeler.AutomationAlgorithm class template where you can define your custom automation algorithm. The template has headers and comments with instructions.

Then, load a data source, create at least one label definition, and on the app toolstrip, In the vision.labeler.AutomationAlgorithm class template that opens, define your custom automation algorithm. Follow the instructions in the header and comments in the class.

Extend Automation Class for Temporal Data or Blocked Images

If the algorithm is time-dependent, that is, has a dependence on the timestamp of execution, your custom automation algorithm must also inherit from the vision.labeler.mixin.Temporal class. Temporal algorithms are useful for labeling videos using the Video Labeler app. For more details on implementing time-dependent, or temporal, algorithms, see Temporal Automation Algorithms.

If the algorithm is blocked image based, your custom automation algorithm must also inherit from the vision.labeler.mixin.BlockedImageAutomation class. A blocked image is a very-large high-resolution image imported in blocks using the Image Labeler app. For more details on implementing blocked image algorithms, see Blocked Image Automation Algorithms.

Implement Core Methods of Automation Algorithm Class

The following table outlines the key methods you implement in a custom automation algorithm class. These methods define how the algorithm behaves during execution within the labeling apps:

Method	Purpose
`checkLabelDefinition`	Validate which label types are compatible with your algorithm
`settingsDialog`	Define custom settings dialog for user interaction (optional)
`checkSetup`	Verify readiness before execution such as any user-setups (optional)
`initialize`	Prepare algorithm state before processing frames
`run`	Core algorithm logic to process each frame
`terminate`	Clean up resources after execution

Create Automation Folder

Create a +vision/+labeler/ folder within a folder that is on the MATLAB path. For example, if the folder /local/MyProject is on the MATLAB path, then create the +vision/+labeler/ folder hierarchy as follows:

projectFolder = fullfile('local','MyProject');
automationFolder = fullfile('+vision','+labeler');
mkdir(projectFolder,automationFolder)

The resulting folder is located at /local/MyProject/+vision/+labeler.

Save Automation Class File to Automation Folder

To use your custom algorithm from within the labeling app, save the file to the +vision/+labeler folder that you created. Make sure that this folder is on the MATLAB search path. To add a folder to the path, use the addpath function.

Refresh Algorithm List in Labeling App

To start using your custom algorithm, refresh the algorithm list so that the algorithm displays in the app. On the app toolstrip, select Select Algorithm > Refresh list.

Import Existing Automation Algorithm in Labeler App

To import an existing custom algorithm into a labeling app, on the app toolstrip, select Select Algorithm > Add Algorithm > Import Algorithm. Refresh the list to make it available.

Custom Automation Algorithm Execution in Labeler App

When you run an automation session in a labeling app, the properties and methods in your automation algorithm class control the behavior of the app.

Check Label Definitions

When you click Automate, the app checks each label definition in the ROI Labels and Scene Labels panes by using the checkLabelDefinition method defined in your custom algorithm. Label definitions that return true are retained for automation. Label definitions that return false are disabled and not included. Use this method to choose a subset of label definitions that are valid for your custom algorithm. For example, if your custom algorithm is a semantic segmentation algorithm, use this method to return false for label definitions that are not of type PixelLabel.

Control Settings

After you select the algorithm, click Automate to start an automation session. Then, click Settings, which enables you to modify custom app settings. To control the Settings options, use the settingsDialog method.

Control Algorithm Execution

When you open an automation algorithm session in the app and then click Run, the app calls the checkSetup method to check if it is ready for execution. If the method returns false, the app does not execute the automation algorithm. If the method returns true, the app calls the initialize method and then the run method on every frame selected for automation. Then, at the end of the automation run, the app calls the terminate method.

The diagram shows this flow of execution for the labeling apps.

Use the checkSetup method to check whether all conditions needed for your custom algorithm are set up correctly. For example, before running the algorithm, check that the scene contains at least one ROI label.
Use the initialize method to initialize the state for your custom algorithm by using the frame.
Use the run method to implement the core of the algorithm that computes and returns labels for each frame.
Use the terminate method to clean up or terminate the state of the automation algorithm after the algorithm runs.