Neural Network Labels
When a neural network is used for object localization, it can classify each localized object and assign it a label — an integer value that indicates which category or class the object belongs to (for example, the color of the object). These labels are visible in the Object Inspector and can be received by your robot controller.
Contents
1 Viewing labels in the Object Inspector
The Object Inspector is available on the Deployment page when you inspect a localized object. It shows the properties of the localized object, including its label when a neural network classification is active.
Note: When the used neural network was not trained to assign labels to localized objects, all objects show the same label value 0 (default).
2 Receiving labels on the robot
Labels are automatically included in the data that are sent back to the robot controller in response to selected action requests. Your robot program can read the label value and use it to decide how to handle each object class.
2.1 Requirements
Labels are sent to the robot only when the robot module version meets the minimum requirement:
Bin Picking Studio: robot module version 1.12 or later
Locator Studio: robot module version 1.5 or later
If your robot module is older than these versions, the label field is not included in the response. Update the robot module to enable label reception.
2.2 Which requests include the label
The label value is included in the response to the following action requests:
Bin Picking Studio (AI Picking solution type)
Trajectory request
Object pose request
Locator Studio (AI Locator solution type)
Get objects request
2.3 What the robot receives
The label is a single integer value appended to the response data for each localized object. The meaning of each integer value depends on the neural network model that was trained for your application - contact your Photoneo Technical consultant for the declaration of the label-to-integer mapping.