Learning Invariant Representations of Objects with State, Pose, and Viewpoint Changes

The core message of this work is to learn object representations that are invariant to changes in the structural form or state of the object, in addition to being invariant to changes in pose and viewpoint.