Hello @t-34400 , thanks a lot for open-sourcing the great repo!

I'm developing a system to track the interaction between users' hand and real objects which means I have to first recognize the objects as well as their location in the environment.

Previously, I was trying to use the depth map (which is 320 x 320) to convert the depth info to real world position. And I found that depth sensor cannot differentiate the height of small objects. For example, if I get a mouse on the table, the point cloud generated from depth map is almost flat.

I am wondering if the Depth images aligned to RGB frames could differentiate the mouse from the table since the repo first generate the 3D model from depth map and then colorize it. Would this step just ignore the models of small objects?

And do you know any way to map the depth map with the RGB file captured by quest3 camera? I found that the depth map captured by depth sensor and the RGB file captured by passthrough camera have different resolutions which is hard to synchronize them.

Any insight would be greatly appreciated, thank you.