How to write a Python function to obtain bbox_2d_loose data based on a existed camera?

Omniverse Archive General Discussion
1

I have a USD file with a pre-built camera, the camera randomize its position in each frame. and I need to access the information of the bounding_box_2d_loose for primitives based on the perspective of the camera to obtain occlusion information for these primitives.

image
image1812×971 85 KB

Here is the code I’ve written, but it doesn’t seem to work.(The annotator returns empty bounding_box_2d_loose data all the time.) Could you please advise on how to rewrite it to achieve my goal?

from pxr import Usd
import omni.usd
import omni.replicator.core as rep

stage = omni.usd.get_context().get_stage()
cam = stage.GetPrimAtPath("/World/Camera")
rp = rep.create.render_product(str(cam.GetPrimPath()), (1024, 1024))

# Acess the data through annotators
bbox_loose = rep.AnnotatorRegistry.get_annotator("bounding_box_2d_loose")
bbox_loose.attach(rp)


async def get_bbox_loose_data():
    global bbox_loose
    await rep.orchestrator.step_async()
    data = bbox_loose.get_data()
    print(data)
    
import asyncio
asyncio.ensure_future(get_bbox_loose_data())