Image-based object recognition in man, monkey and machine

Fitting objects into holes: On the development of spatial cognition skills.

“…The critical difference between the three tasks in Experiment 2 was that the base shape delimited a negative space in the hole-fitting task and a positive space in the tower and formfitting tasks. Besides, we have seen that it is easier to work with familiar objects presented in the same way than with objects that look different (Rudge and Warrington 1991;Skouteris, McKenzie et al 1992;Tarr and Bülthoff 1995;Tarr and Bülthoff 1998). While in the tower condition the two objects were copies of one another, the other tasks had objects that required a rendering of a 3D object to a 2D display.…”

“…Most likely, this has to do with the fact that it is not the whole 3D object that is represented by the aperture, but just one of its cross-sections. The children see the object from above showing its short side, they might relate it to the aperture that shows a matching silhouette from the viewpoint parallel to the object's major axis (Rock 1974;Tarr and Bülthoff 1995;Tarr and Bülthoff 1998). If the object is lying down, the children get quite a different picture than the opening of the aperture, and will most likely produce a worse result.…”

“…If the 2D and 3D figures are not rotated in different ways but are presented in the same direction to the child so that their appearance looks similar, it is easier to relate them to each other. Also, the more experience the child has had with objects of a certain class, the easier it is to recognize a similar object (Tarr and Bülthoff 1995;Tarr and Bülthoff 1998).…”

“…Thus, it is only logical to distinguish an object presented from a familiar angle prior to one presented from a new angle (Piaget 1952;Rudge and Warrington 1991;Tarr and Bülthoff 1995;Tarr and Bülthoff 1998).…”

“…For example, a regular cylinder has a circular silhouette from a viewpoint parallel to its major axis, a rectangular silhouette from a viewpoint perpendicular to its major axis, and a family of silhouettes of more complex shapes from other viewpoints. It is not surprising that object recognition is easier when the perspectives are similar (Rudge and Warrington 1991;Tarr and Bülthoff 1995;Tarr and Bülthoff 1998). It should therefore be easier to relate the horizontal 2D silhouette to a standing object than a lying one.…”