Coordinate Frames in MobilityGen Depth Images #29
Description
Dear MobilityGen Team,
I'm currently trying to implement MobilityGen for my quadruped robot. I had a question regarding the coordinate frames used for the exported depth images:
Are the depth images encoded in the camera frame, or are they transformed into a fixed frame (e.g., world or robot base)?
If they are in the camera frame, is the camera pose (e.g., transformation from world to camera) provided per frame, so that the depth data can be transformed into a consistent global?
Do you have any recommendations for using the depth images when training terrain-aware locomotion policies, particularly when a consistent spatial reference is required?
Thank you for your time, and for the amazing work.
Best regards,