ROS Resources: Documentation | Support | Discussion Forum | Index | Service Status | ros @ Robotics Stack Exchange
Ask Your Question

Revision history [back]

click to hide/show revision 1
initial version

A semantic map is a data structure, it is not a processing unit, so there are no inputs or outputs. The map consists of object instances at different poses in space and their composition of sub-parts. The objects are described as instances of more generic object classes and inherit their properties. I've already pointed you to related publications and can only recommend to read about the concepts behind semantic maps, e.g. in [1].

You can, in principle, create semantic maps from very different input sources: In [1], RGB-D data is used, in the KnowRob tutorial at the ROS Fall School 2010, we have interfaced the tabletop_object_detector, you can use image-based recognition methods like the RoboEarth object recognition system, or you can build the map yourself, manually or with the (experimental) semantic map editor (some docs).

Moritz

[1] http://ias.cs.tum.edu/_media/spezial/bib/iros12semantic_mapping.pdf