Direct grounding of spatial relations that correspond to most commonly used in language spatial prepositions and abstract action representations based on the temporal evolution of object-wise spatial relations.
Spatial Relations/demo.m for quick examples on how to compute PVS-based action descriptors given tracked object point clouds and how to compare action descriptors.
Computing distances between PVS-based action descriptors is formulated as a MIQP (Mixed-Integer Quadratic Program) and the current implementation (
compareActions.m) relies on OPTI Toolbox (SCIP solver). Switching to another MIQP solver should not be difficult (the laborious part of building the objective/constraint matrices is done).
This is the original implementation for the work introduced in the ICRA 2015 paper:
Learning the Spatial Semantics of Manipulation Actions through Preposition Grounding [pdf]
Konstantinos Zampogiannis, Yezhou Yang, Cornelia Fermüller, Yiannis Aloimonos