Semantic Segmentation labels not behaving as expected

[rosalie_va]

Hi! I have a question: I want to retrieve the following data with semantic segmentation (with Replicator of course):

{'0': {'class': 'UNLABELLED'}, '4': {'class': 'wheel'}, '1': {}, '3': {'class': 'case'}, '2': {'class': 'handle'}}

But instead when capturing the semantic segmentation it is labelled like so:

{'(0, 0, 0, 0)': {'class': 'UNLABELLED'}, '(140, 25, 255, 255)': {'class': 'wheel'}, '(25, 255, 82, 255)': {}, '(255, 197, 25, 255)': {'class': 'case'}, '(140, 255, 25, 255)': {'class': 'handle'}}

Does someone know what’s happening and how to fix it? I once managed to get it to work and I vaguely recall that I fixed it by adding a Ground Plane to the scene, but that doesn’t work anymore…

[ahaidu]

Can you provide a small repro code snippet on how you are getting the results?

[rthaker]

Hi @rosalie_va - Are you still having this issue? If yes, then please share the requested info in the earlier post.

[ahaidu]

The annotators have an updated documentation, can you check if the described outputs solve your issue?

Annotators Information — Omniverse Extensions documentation (nvidia.com)