Bounding Box Select Agent
The BoundingBoxSelectAgent detects an object in an image and selects its bounding box.
Initialization
Parameters
The BoundingBoxSelectAgent is initialized with two arguments:
BoundingBoxSelectAgent(classes, model(Optional))
Specifies the list of objects to detect within an image for the object detection task.
Each object of interest (or “class”) should be included within the classes list.
Represents the model used to perform the detection task. If left unspecified, this parameter defaults to GroundingDINO().
The supported BoundingBoxModel models can be found below:
Note: Since a text embedding of each class will be used to identify objects, a more specific class name will yield a more specific detection. You should scope your class names accordingly.
For instance, a
"car"class will yield more general detections than a"blue Honda Accord"class.
Example
BoundingBoxSelectAgent(classes=["person"])
The output from this agent will be an ExecutionNode containing the previous image and the detection results.