Please enable JavaScript.
Coggle requires JavaScript to display documents.
Semantic Slam Problems (Current SLAM techniques struggle (to maintain…
Semantic Slam Problems
Object Detection
SLAM
Semantic Representation
Data Association
Optimization
Evaluation
More meaningful maps
The limitations of purely geometric maps have been identified
Object SLAM
Bounding box (course priors)
Constrained ellipsoids from 2D boxes
3D model matching (rich priors)
More accurate maps
SotA geometric SLAM methods struggle in
Highly dynamic environments
indirect slam struggles in low feature environments
environments that change over time
More robust maps
Data association
Current SLAM techniques struggle
to maintain accuracy
in highly dynamic environments
with fast agent motion
when environments change over time
to remain robust
to be meaningful