The visual active memory perspective on integrated recognition systems

摘要: Object recognition is the ability of a system to relate visual stimuli its knowledge world. Although humans perform this task effortlessly and without thinking about it, general algorithmic solution has not yet been found. Recently, shift from devising isolated techniques towards integrated systems could be observed [Y. Aloimonos, Active vision revisited, in: Y. Aloimonos (Ed.), Perception, Lawrence Efibaum, 1993, pp. 1-18; H. Christensen, Cognitive (vision) systems, ERCIM News (April, 2003). 17-18]. The active memory (VAM) perspective refines view an interactive computational framework for in human everyday environments. VAM line with recently emerged Vision paradigm [H. 17-18] which concerned that evaluate, gather integrate contextual analysis. It consists processes generate by means tight cooperation perception, reasoning, learning prior models. In addition, emphasizes dynamic representation gathered knowledge. assumed structured hierarchy successive mediate modularly defined processing components system. Recognition take place stress field objects, actions, activities, scene context, user interaction. paper, we exemplify existing demonstrator systems. Assuming three different perspectives (biological foundation, engineering, computer vision), will show concept central cognitive capabilities it leads more object framework.