Using camera images as sensor tensors to train RL polices

[scode-cmd]

Are there any updates on some official examples for camera input to train RL policies? I was confused watching the "Isaac Gym: The Next Generation" talk which gives the impression that it's not possible to use camera images as sensor tensors to train RL polices using Isaac Gym.

Is this still the case?

https://www.nvidia.com/en-us/on-demand/session/gtcspring22-s41582/?playlistId=playList-63b157fe-95fe-4b93-8b9b-d731be32ec29

time stamp 31:49