How to create a pointcloud from a sequence of RGBD images with semantic information

Hi guys,
first of all thanks for the great work! I’m looking for a solution for a problem related to pointclouds from RGBD and odometry integration. I have no problem in generating the pointcloud from rgbd data, but I was wondering if exists a method to “attach” more information (e.g., semantic class and so on) to each voxel. I couldn’t really find a way. Do you have any suggestions? I’m new to the topic, any suggestions would be really appreciated.
Thanks!