Graphical Models for Machine Perception
Many difficult machine perception problems, from stereo vision to terrain classification, can be formulated in terms of graphical models. These formulations are appealing: they are often conceptually simple yet produce excellent results. They can also handle both noisy and missing data. While the models are easy to understand, computing solutions can be expensive. This talk will describe how several robotics perception problems can be posed in terms of graphical models and the techniques that make them tractable.