Using ZED SDK's provided object detection API alongside a custom model

Software
1

I’m looking at using the provided ZED object detection model alongside my own custom trained model. Currently with how the SDK works, you can only have one or the other. I’ve found the provided model to be robust enough for my testing.

Is there any plans to change how the object detection API works so I can get the ZED camera to give me both it’s detections and my own detections.

At the moment you must select between one of the two

image

2

You can perform concurrent inference by using different instance_module_id values when calling enableObjectDetection:

You can find examples here:

1 Like