[Question] Can visualization markers retain the markers displayed at each step like debug_draw and provide corresponding methods to clear existing markers?

[amatsukaze233]

Question

Currently, I need to visualize at least 60,000 and increasing sphere markers and be able to record RGB information with a camera for reinforcement learning observation.

When using visualization markers, all marker coordinates need to be inputted at each step, and as the number of markers increases, the performance loss becomes greater.

When using debug_draw, only the new marker coordinates need to be inputted at each step, making rendering relatively faster.

However, the visual effect seen through the camera is very poor when using debug_draw.

I didn't quite understand the usage of point instancer in the source code.

[RandomOakForest]

Thank you for posting this. Could you help us reproduce this issue? could you share the relevant code you are using for this? Thanks.