KARMIC TABULA RASA
In this paper, we hypothesize a two-step process to visual perception and recognition of objects. The first level is based on an evolutionary cognitive algorithmic process which we label as â€˜karmicâ€™ (k). The second level depends on repeated learning of correlated local features in the object space, which we call the â€˜tabula rasaâ€™ (TR). We also present models for â€˜kâ€™ and â€˜TRâ€™ features grounded in our theory of Recognition by Component Affordances (RBCA). Using the combined features, which we call k-TRONs composed of 25 structural affordances and augmented by 10 material affordances, we support recognition of over 200 categories of commonly occurring objects. We argue that k-TRONs form the basis of evolutionary object recognition and present psychophysical and neurobiological evidence to support our hypothesis. Allied aspects of recognition such as Novelty detection, Equivalence classes, Recognition of articulated and natural objects, Attention, Saliency, Memory and Scale of analysis are also addressed.