Modules in vision: A case study of interdisciplinarity in cognitive science

Abstract

The usefulness of interdisciplinarity in cognitive science (CS) is demonstrated by an overview of recent algorithms for recovering 3-D aspects from a 2-D input. The recovery processes are considered to be modular in Fodor's (1983) sense. This tenet, based on a thesis from philosophy of mind (PhM), proves to have serious impact both on computer vision (CV) as such and on computational theories of perception (CTP), which provide a compromise between classic indirect and direct theories (ITP and DTP) in perceptual psychology (PP). Both existent and possible mutual interchanges between CV, PP, and PhM are specified in current and future research on modular recovery processes such as shape from shading, depth from stereo, and structure from motion. Also, attention is paid to (meta) criticisms of PP and PhM on concrete hypotheses of CTP and on the CTP approach as a whole. For example, the relative independence of modular low-level vision processes is questioned, and the lack of an explanation of intentionality is highlighted. A plea is made to attempt to solve these fundamental (meta)criticisms within the CTP paradigm, since there are no logical arguments against this possibility, and because recent CTP theories are tackling these problems seriously.