Scaleable object recognition with a belief model
First Claim
1. A system operative to recognize objects in content comprising:
- a blackboard comprising a plurality of experts, and data comprising original input data and data created by processing of any of said plurality of experts, and a controller operative to control said experts;
a belief model, coupled to said controller, comprising a set of beliefs and probabilities associated with each belief of said set of beliefs;
a belief network, coupled to said controller; and
a relations subsystem, coupled to said controller.
8 Assignments
0 Petitions
Accused Products
Abstract
An exemplary embodiment of the present invention is directed to a system, method and computer program product for providing an object recognition blackboard system architecture. The system for recognizing objects in content can include: a blackboard comprising a plurality of experts, and data comprising original input data and data created by processing of any of the plurality of experts, and a controller operative to control the experts; a belief model, coupled to the controller, comprising a set of beliefs and probabilities associated with each belief of the set of beliefs; a belief network, coupled to the controller; and a relations subsystem, coupled to the controller.
-
Citations
23 Claims
-
1. A system operative to recognize objects in content comprising:
-
a blackboard comprising a plurality of experts, and data comprising original input data and data created by processing of any of said plurality of experts, and a controller operative to control said experts;
a belief model, coupled to said controller, comprising a set of beliefs and probabilities associated with each belief of said set of beliefs;
a belief network, coupled to said controller; and
a relations subsystem, coupled to said controller. - View Dependent Claims (2, 3, 4, 5, 6, 7, 8, 9, 10, 11, 12, 13, 14, 15, 16, 17, 18, 19, 20, 21)
-
-
22. A method of recognizing objects comprising:
-
identifying classes of objects specified by a user using a plurality of cooperative object recognition experts;
achieving higher accuracy from using in parallel said plurality of cooperative object recognition experts than is achievable using in serial said plurality of cooperative object recognition experts;
supporting scaleability of performance including supporting multiple processors;
developing a belief model including specifying specified associations among said objects, learning learned associations among said objects, representing said specified and learned associations, and forming a belief network wherein said belief network is at least one of a Bayesian Network and a Dempster Shafer Network; and
deducing shadow objects from said belief model.
-
-
23. A method for adding a new expert to a blackboard comprising:
-
creating an expert;
encapsulating said expert;
compiling said expert;
adding a stub function to a blackboard;
determining if output of said expert is new and if new, then adding the output'"'"'s class to said blackboard, and updating a belief model by providing truth data file data to a learning system; and
creating a rule to control when said new expert is to be executed when supporting evidence is found to exceed an adaptable threshold.
-
Specification