Techreport,
Distributed Probabilistic Learning for Camera Networks
DCS-TR-696. Rutgers, (June 2012)
Abstract
Probabilistic approaches to computer vision typically assume a centralized setting, with the algorithm granted access to all observed data points. However, many problems in wide-area surveillance can bene t from distributed modeling, either because of physical or computations constraints. In this work we present an approach to estimation and learning of generative probabilistic models in a distributed context. In
particular, we show how traditional centralized models, such as probabilistic principal component analysis (PPCA), can be learned when the data is distributed across a network of sensors. We demonstrate the utility of this approach on the problem of distributed ane structure from motion (SfM). Our experiments suggest that the accuracy of the accuracy of the learned probabilistic structure and motion models rivals that of traditional centralized factorization methods.
