I’m going to use Isaac Gym for vision-based RL tasks, but the speed of rendering images or point clouds is definitely the bottleneck, so the whole pipeline is very slow and even can not run in a reasonable time!
Have you found the solution? I have the same question .
Interested, It is on my watch too.
Meanwhile do you have a code snippet using camera sensor ? Can’t find any in OmniIsaacgym.