Hello. According to the scheme, presented in the paper, and the code investigation, loop closure results from RPGO are just published outside Kimera. However it looks strange, because a new map after drift correction (last keyframes and landmarks) should be used by the backend for poses estimation (like in ORB-SLAM 2/3). Have I missed something?
Hello. According to the scheme, presented in the paper, and the code investigation, loop closure results from RPGO are just published outside Kimera. However it looks strange, because a new map after drift correction (last keyframes and landmarks) should be used by the backend for poses estimation (like in ORB-SLAM 2/3). Have I missed something?