PSSC: Pictorial Structures for Scene Context

We propose to develop a Visual Context Understanding (Vis-CU) intelligent agent.Vis-CU tightly integrates vision processing with cognitive reasoning.The vision processing draws the wide array of vision processing capabilities, integrated through the University of Michigans sparse component model into the Soar Cognitive Architecture (Soar).This tight integration facilitates the development of a rich, layered understanding of context within an architecture that serves as a general purpose reasoning framework.This approach enables a wide variety of reasoning techniques to be applied to scene understanding problems, particularly enabling systems to adapt to real-world contexts.By quickly adapting processing based on the current context, systems become far more robust in the face of real-world complexity.