Understanding objects in terms of their individual parts is important,
because it enables a precise understanding of the objects' geometrical
structure, and enhances object recognition when the object is seen in a novel
pose or under partial occlusion. However, the manual annotation of