Levitt, T., Mullin, J., and Binford, T., "Model-based Influence Diagrams for Machine Vision," In Proceedings of the Fifth Workshop on Uncertainty in Artificial Intelligence, Associationfor Uncertainty in Artificial Intelligence, Mountain View, California, 1989, pp. 233-244.  Home/Search   Document Not in Database   Summary   Related Articles   Check

....from their rear tail light assemblies. The extended [IG01] provides a survey of the related work, placing our work in the context of related frameworks for building [PL94] and analysing [RH92] vision system; and of related approaches, such as information theory [GJ96] and influence diagrams [LAB90] It also discusses other applications of our framework in particular, showing that it can be used when recognizing faces, within the modern eigenvector approach. 2 Framework 2.1 Standard Approaches There are many approaches to recognition. A strictly bottom up approach performs a series ....

....scene contains a person, etc. These systems differ from strictly bottom up schemes by then switching to a topdown mode: given sufficient evidence to support one interpretation, they seek scene elements that correspond to the parts of the proposed real world object that have yet to be found [LAB90] Notice the model based systems have more prior knowledge of the scene contents than the strictly bottom up schemes in particular, they have some notion of models which they exploit to be more efficient. We propose going one step further, by using additional prior knowledge to ....

T S Levitt, J M Agosta, and T O Binford. Model-based influence diagrams for machine vision. In UAI, 1990.

....scene contains a person, etc. These systems differ from strictly bottom up schemes by then switching to a top down mode: given sufficient evidence to support one interpretation, they seek scene elements that correspond to the parts of the proposed real world object that have yet to be found [LAB90] Notice the model based systems have more prior knowledge of the scene contents than the strictly bottom up schemes in particular, they have some notion of models which they can exploit to be more efficient. We propose going one step further, by using additional prior knowledge to ....

....approaches to face recognition [TP91; PMS94; PWHR98; EC97] none explicitly address the issues of efficiency. We used one of the popular and successful methods as the basis to our approach and performed a systematic study of the various efficiency and accuracy related issues. Levitt et al. LAB90; BLM89; LHD 93] have applied Bayesian inference methods and influence diagrams to image interpretation. We however provide a way to adjust the optimization function. Our work also further motivates the use of maximum expected utility in such systems. As our system is seeking a policy ....

T S Levitt, J M Agosta, and T O Binford. Model-based influence diagrams for machine vision. In UAI, 1990.

.... the methods presuppose the availability of perceptual detectors for cues of collaborative action that subsequently bias a logical inference process toward correct interpretations [22] Bayesian networks, also known as belief networks, have been used for visual recognition of static objects (e.g. [27] and others) and for visual attention selection (e.g. 43] Promising work on recognizing single agent action from trajectory information using transition diagrams and fuzzy reasoning [30] led us to investigate the use of belief networks for multiagent action recognition, which more explicitly ....

T. S. Levitt, J. M. Agosta, and T. O. Binford, Model-based influence diagrams for machine vision, in Uncertainty in Artificial Intelligence (M. Henrion, R. D. Shachter, L. N. Kanal, and J. F. Lemmer, Eds.), Vol. 5, pp. 371--388, Elsevier Science (North Holland), Amsterdam, 1990.

.... is observable given the presence of p; and P (p) and P (e 1 ; e 2 ; e n ) are the probabilities of p and the evidence occurring (a priori probabilities) The conditioning of the bodies of evidence is frequently used in conjunction with a directed, acyclic graph model of the percept [2,28,29,42]. Nodes show PART OF relationships between the component features of a percept, and observations of these features contribute the evidence used by the inference process. These graphical representations are generally referred to as Bayesian belief nets and or influence diagrams. While Bayesian ....

Levitt, T.S., Agosta, J.M., and Binford, T.O., "Model-Based Influence Diagrams for Machine Vision", Uncertainty in Artificial Intelligence 5, M. Henrion, R.D. Shacter, L.N. Kanal, and J.F. Lemmer, ed., Elsevier Science Publishers, B.V., 1990, pp. 371-388.

.... the methods presuppose the availability of perceptual detectors for cues of collaborative action that subsequently bias a logical inference process towards correct interpretations [22] Bayesian networks, also known as belief networks, have been used for visual recognition of static objects (e.g. [27] and others) and for visual attention selection (e.g. 43] Promising work on recognizing single agent action from trajectory information using transition diagrams and fuzzy reasoning [30] led us to investigate the use of belief networks for multi agent action recognition, which more explicitly ....

T.S. Levitt, J.M. Agosta, and T.O. Binford. Model-based influence diagrams for machine vision. In M. Henrion, R.D. Shachter, L.N. Kanal, and J.F. Lemmer, editors, Uncertainty in Artificial Intelligence, volume 5, pages 371--388. Elsevier Science Publishers B.V. (North Holland), 1990.

....demonstrate its practicality. 1 INTRODUCTION Over the last several years Bayesian networks have been applied to a wide variety of problems ranging from medical diagnosis [ Heckerman et al. 1992; Horvitz et al. 1988 ] and natural language understanding [ Charniak and Goldman, 1991 ] to vision [ Levitt et al. 1989 ] and map learning [ Dean, 1990 ] A central problem in such applications is to use the network to generate explanations for observed data. Such explanations correspond to instantiations of the network (i.e. value assignments to each node in the network) with the structure of the network ....

T. Levitt, J. Mullin, and T. Binford. Model-based influence diagrams for machine vision. In Proceedings of CUAI-89, pages 233--244, 1989.

....among multiple goals of multiple agents. Search based systems designed specifically to recognize multi agent goals and actions outside of probabilistic frameworks [28, 3, 31] are sensitive to noisy data and detectors. Belief networks have been used for visual recognition of static objects [20, 1] and for visual attention selection[29] Promising work on recognizing single agent action from trajectory information using transition diagrams and fuzzy reasoning [19, 23] led us to investigate the use of belief networks, which more explicitly represent knowledge dependencies and are ....

T.S. Levitt, J.M. Agosta, and T.O. Binford. Modelbased influence diagrams for machine vision. In M. Henrion, R.D. Shachter, L.N. Kanal, and J.F. Lemmer, editors, Uncertainty in Artificial Intelligence, volume 5, pages 371--388. Elsevier Science Publishers B.V. (North Holland), 1990.

....approximate inference algorithms available for performing belief updating [16] in this paper we are not concerned with the particular algorithm. 3. 2 Why Dynamic Belief Networks Belief networks have been been used in various applications, such as medical diagnosis [20] and model based vision [12], which initially were more static, i.e. essentially the nodes and links do not change over time. Such approaches involve determining the structure of the network; supplying the prior probabilities for root nodes and conditional probabilities for other nodes; adding or retracting evidence about ....

T.S. Levitt, J. M. Agosta, and T.O. Binford. Model-based influence diagrams for machine vision. In Proc. of the Fifth Workshop on Uncertainty in Artificial Intelligence, pages 233--244, 1989.

.... form of a Bayes net, designed for modeling a taxonomic hierarchy, is applied to image labeling in [Chou and Brown, 1990] The first large experimental system that uses Bayes nets for computer vision is [Levitt, 1986; Levitt et al. 1989] and that work has continued, e.g. Agosta, 1990; Levitt et al. 1990; Mann and Binford, 1992; Agosta, 1991; Chelberg, 1989] Bayes nets have also been applied to mobile robots [Dean et al. 1990; Dean and Wellman, 1991; Dean and Kirman, 1992] Other vision applications of Bayes nets include [Anderson et al. 1989; Jensen et al. 1992; Sarkar and Boyer, 1992; ....

T. S. Levitt, J. M. Agosta, and T. O. Binford, "Model-Based Influence Diagrams for Machine Vision," In Uncertainty in AI 5, pages 233--244. North-Holland, 1990.

.... Model Construction Bayesian or Belief networks [29] 1 are graphical probabilistic dependency models which integrate a mechanism for inference under uncertainty with a secure Bayesian foundation, were initially applied to static domains, where the network structure does not change over time [33, 25]. These approaches involved determining the network structure, supplying prior and conditional probabilities, adding or retracting evidence and repeating the inference algorithm for each change in the evidence. The complexity and size of networks for domains such as natural language understanding ....

T. Levitt, J. M. Agosta, and T. Binford, "Model-based influence diagrams for machine vision," in Proc. of the Fifth Workshop on Uncertainty in Artificial Intelligence, pp. 233--244, 1989.

....interpretation. This kind of explicit reasoning about constraints is unable to deliver the efficient performance required for visual surveillance of dynamic scenes. The computational efficiency of such static vision systems can be improved by applying Bayesian belief networks and decision theories [6, 28, 29]. However, they have not yet addressed the specific computational difficulties involved in the interpretation of image sequences. On the other hand, work using model based approaches in image sequence analysis by Koller et al., Lowe, and Worrall et al. [24, 30, 51] does effectively address the ....

T.S. Levitt, J.M. Agosta, and T.O. Binford. "Model-based influence diagrams for machine vision ". In Uncertainty in Artificial Intelligence 5. North-Holland, 1990.

....a few assumptions. 2 Related Work 2. 1 Bayesian Networks in Image Understanding Bayesian networks, in combination with geometric reasoning systems, have been used in 3 D object understanding to relate model components to predicted appearances and to control the interpretation process ( 16] 3] [15], 2] 19] 18] 23] The highest level of the model is a hierarchy of 3 D components, which predict the appearance of surfaces and contours. These surfaces and contours in turn predict the appearance of edges. Matching edges, finally, form the evidence nodes. Another application is the ....

T.S. Levitt, J.M. Agosta, and T.O. Binford. Model-based influence diagrams for machine vision. Uncertainty in AI, 5:233--244, 1990.

....representations used in computer vision. Traditional techniques typically attempt to model a percept in terms of component features. The composition of the percept is expressed with some variant of a directed, acyclic graph (DAG) most notably Bayesian belief nets [ 2 ] influence diagrams [ 7 ] , or relational graphs [ 4, 5 ] The desired percept is the root of a graph with component features as the children. DAG models exploit the advantages associated with graphs: compact storage, ease of traversal, and the ability to append information (such as evidential contribution) to vertices. ....

....Form using prefix notation in Figure 1. Observational Grammar G p must be supplemented because it suffers from the same difficulties as traditional DAG based methods in representing the transformation of evidence for features in a model to evidence for a percept. DAG based techniques such as [ 2, 7 ] attempt to embed the evidential contribution of a feature to the overall belief in a percept directly at each feature node in the DAG. This is challenging because the evidential contribution of a feature may depend on the context. The description of the sensing objective affects the belief in the ....

Levitt, T.S., Agosta, J.M., and Binford, T.O., "Model-Based Influence Diagrams for Machine Vision ", Uncertainty in Artificial Intelligence 5, M. Henrion, R.D. Shacter, L.N. Kanal, and J.F. Lemmer, ed., Elsevier Science Publishers, B.V., 1990, pp. 371-388.

.... are being increasingly used in various real world applications, including medical diagnosis (Suermondt Amylon, 1989; Andreassen et al. 1987) telecommunications (Ezawa Norton, 1995) information retrieval (Fung Favero, 1995) system troubleshooting (Heckerman et al. 1994a) vision (Levitt et al. 1989) and language understanding (Charniak Goldman, 1989) This paper presents K2 AS, a novel method for inducing selective Bayesian networks, and compares its classification performance with three other probabilistic induction methods: a Bayesian network induction method that uses all features ....

Levitt, T., Mullin, J., & Binford, T. (1989). Model-based influence diagrams for machine vision.

....label has a corresponding probability, that probability can be used to rank order candidate image faces for search. In selecting which recovered face to focus our attention on, we utilize a decision theoretic approach using a Bayesian framework. A similar approach was reported by Levitt et al. [28], who use Bayesian networks for both model representation and description of recovered image features. Specifically, they use Bayesian networks for both data aggregation and selection of actions and feature detectors based on expected utility. The approach is thus centered around the use of a ....

T. Levitt, J. Agosta, and T. Binford. Model based influence diagrams for machine vision. In M. Herion, R. Shacter, L. Kanal, and J. lemmer, editors, Uncertainty in Artificial Intelligence 5, volume 10 of Machine Intelligence and Pattern Recognition Series, pages 371--388. North Holland, 1990.

.... in the diagnosis of medical patients [Heckerman, 1991, Andersen et al. 1989, Heckerman et al. 1990, Peng and Reggia, 1990] and malfunctioning systems [Agogino et al. 1988] to understand stories [Charniak and Goldman, 1991] to filter documents [Turtle and Croft, 1991] to interpret pictures [Levitt et al. 1990], to perform filtering, smoothing, and prediction [Abramson, 1991] to facilitate planning in uncertain environments [Dean and Wellman, 1991] and to study causation, nonmonotonicity, action, change, and attention. Some of these applications are described in a tutorial article by [Charniak, 1991] ....

T.S. Levitt, J.M. Agosta, and T.O. Binford. Model-based influence diagrams for machine vision. In M. Henrion, R.D. Shachter, L.N. Kanal, and J.F. Lemmer, editors, Uncertainty in Artificial Intelligence 5, pages 371--388. North Holland, Amsterdam, 1990.

....No. F30602 91 C 0041. 1 Introduction Belief Networks [Pearl, 1988] which integrate a mechanism for inference under uncertainty with a secure Bayesian foundation, were initially applied to fairly static domains, where the nodes and arcs do not change over time [Spiegelhalter et al. 1989, Levitt et al. 1989]. These approaches involved determining the network structure, supplying prior and conditional probabilities, adding or retracting evidence and repeating the inference algorithm for each change in the evidence. The complexity and size of networks for domains such as Natural Language Understanding ....

Levitt, T.S.; Agosta, J. M.; and Binford, T.O. 1989. Model-based influence diagrams for machine vision. In Proc. of the Fifth Workshop on Uncertainty in Artificial Intelligence. 233--244.

....can readily handle incomplete information. This evidence is propagated through the network affecting the overall joint distribution (as represented by the conditional probabilities) Belief networks have been been used in various applications, such as medical diagnosis [18] and model based vision [11], which initially were more static, i.e. essentially the nodes and links do not change over time. Such approaches involve determining the structure of the network; supplying the prior probabilities for root nodes and conditional probabilities for other nodes; adding or retracting evidence about ....

T.S. Levitt, J. M. Agosta, and T.O. Binford. Model-based influence diagrams for machine vision. In Proc. of the Fifth Workshop on Uncertainty in Artificial Intelligence, pages 233--244, 1989.

....1986, and a projection from 1987 through 2005. The system included political, social and economic variables, most of which were chance nodes with numeric or linguistic states. Scenarios were tested by using Pearls methods for data fusion and propagation and for stochastic simulation [110] IES [13, 92, 93, 18, 90]: A system for machine vision, radar signal detection, and military reconnaissance. IES operates by dynamically and incrementally forming a hierarchy of IDs for image interpretation; it attempts to use visual information as a way of understanding the movements of troops and other objects. Earlier ....

T.S. Levitt, J.M. Agosta, and T.O. Binford. Model-Based Influence Diagrams for Machine Vision. In L. Kanal, M. Henrion, R. Shachter, and J. Lemmer, editors, Uncertainty in Artificial Intelligence 5, pages 371--388. North Holland, 1990.

....the present paper is a clarification and further development of the points made here. Several other groups are working with the use of Bayesian approaches to computer vision. In particular, the group from the Robotics Laboratory at Stanford University is working along the same lines as we are [ Levitt et al. 1989; Binford et al. 1989 ] However, while they are attempting to build a vision system from top to bottom as an influence diagram [ Shachter, 1988 ] we are considering the CPN as part of one agent in a multi agent system, and our principal interest is how such a Bayesian agent may interface ....

T. S. Levitt, J. M. Agosta, and T. O. Binford. Model-based influence diagrams for machine vision. In Proceedings from the Fifth Workshop on Uncertainty in Artificial Intelligence, pages 233 -- 244, 1989.

No context found.

Levitt, T., Mullin, J., and Binford, T., "Model-based Influence Diagrams for Machine Vision," In Proceedings of the Fifth Workshop on Uncertainty in Artificial Intelligence, Associationfor Uncertainty in Artificial Intelligence, Mountain View, California, 1989, pp. 233-244.

No context found.

T. S. Levitt, J. M. Agosta, and T. O. Binford (1989), Model-Based influence diagrams for machine vision, Proceedings of the Fifth Conference on Uncertainty in Artificial Intelligence, pp. 233-244.

No context found.

T. S. Levitt, J. M. Agosta, and T. O. Binford (1989), Model-Based influence diagrams for machine vision, Proceedings of the Fifth Conference on Uncertainty in Artificial Intelligence, pp. 233-244.

No context found.

Levitt, T.S., Agosta, J.M., and Binford, T.O., "Model-Based Influence Diagrams for Machine Vision", Uncertainty in Artificial Intelligence 5, M. Henrion, R.D. Shacter, L.N. Kanal, and J.F. Lemmer, ed., Elsevier Science Publishers, B.V., 1990, pp. 371388.

Online articles have much greater impact   More about CiteSeer.IST   Add search form to your site   Submit documents   Feedback  

CiteSeer.IST - Copyright Penn State and NEC