Closed farhanmubasher closed 6 years ago
DSO is "only" a keyframe odometry. There is no persistent map so far.
We have released LDSO, which integrates loop closure detection and pose graph optimization. This is still not quite what you need for map re-use. The missing parts are storing a map and releasing it and registering a newly initialized odometry against the old map (could be done with the same loop closure detection). Eventually one also would need some map management / keyframe pruning for long term operation, but all the building blocks should be there.
Here is the link: https://github.com/tum-vision/LDSO
I'll close this issue for now.
Hi all.. I'm using dso_ros for real time tracking and i'm getting poses from camToWorld.data() ( e.g x,y,z, qx, qy, qz, qw). But how can i build map using these poses and reuse that map. For exmaple, I want to move a robot in a square. So firstly i will move it manually in a square so that it can build a map and after that it should move in that map autonomously .
is there any other way to do it ? (Build map from unknown environment using dso and then reuse it so robot can move autonomously in that involvement). Thanks in advance !!