A Knowledge Framework For Seeing and Learning
Paul R. Cooper, Matthew A Brand · 1997
Abstract Seeing requires knowledge—lots of knowledge. While this has been understood clearly from the beginnings of computer vision, progress in building knowledge intensive vision systems has been slow, if not absent entirely. There are a number of reasons for this, including serious problems with: • simply identifying knowledge that is useful for vision, • understanding how to build vision systems that can exploit high-level knowledge, including understanding how to build vision systems that can even so much as bridge levels of abstraction, • understanding how to generalize tha applicability of knowledge beyond narrow domains.