MIT-SPARK/Kimera-VIO

Pure localization in the voxblox map created by Kimera-Semantics

madhu-korada opened this issue · 2 comments

I am able to generate a 3d semantic mesh (.ply) file using Realsense using Kimera VIO and Kimera Semantics. I am not sure if this issue belongs here, but I have few questions.

  1. I am not sure how can I do only localization in the map that I created. Currently everytime I run the kimera pipeline it generates the map from scratch, is there a way I can use the map that I have already built (like only localization mode in RTAB-Map)?
  2. Is there any way I can extract local map while localization? Per frame mesh from Kimera-Mesher is present but it doesn't seem accurate enough.

Any help or suggestions would be helpful. Thanks in advance.

  1. We don't at this time have relocalization tools in Kimera-VIO or Kimera-Semantics, so this would be a bit of work to get going.
  2. We don't have any API to save the Kimera-Mesher per-frame-mesh to disk as it's not accurate enough for any real use, as you noted. It's meant more for fast applications (e.g., obstacle avoidance). The dense global map can be saved using voxblox tools; there is a rosservice you can call from command line or programmatically. This will not be the local map though. Example:
rosservice call /kimera_semantics_node/generate_mesh "{}"

Accidental close