University of Cambridge > > Inference Group > Bayesian Learning for Visual Inference

Bayesian Learning for Visual Inference

Add to your list(s) Download to your calendar using vCal

If you have a question about this talk, please contact Phil Cowans.

Every second, gigabytes of data arrive at our eyes, yet our brain effortlessly translates this into concise descriptions of the world enabling us to perform everyday tasks. As information engineers, our responsibility is to manage the overwhelming quantities of information available to us and in my research I have taken inspiration from humans to learn the mappings between high-dimensional image data and a problem-specific output space. Such mappings are learnt discriminatively from a set of labelled training data using the Bayesian rules of inference to pragmatically account for uncertainty, incorporate prior knowledge and set parameter values. The benefits of learning mappings (as opposed to defining a model of image generation) are efficiency and the ability to generalize when images change in some previously unseen way. I will demonstrate how this general concept has been used to create a system for tracking moving objects in video sequences and to create a one-dimensional visual interface that can be used to drive the Dasher typing system. This is joint work with Andrew Blake (Microsoft Research) and Roberto Cipolla (University of Cambridge).

This talk is part of the Inference Group series.

Tell a friend about this talk:

This talk is included in these lists:

Note that ex-directory lists are not shown.


© 2006-2024, University of Cambridge. Contact Us | Help and Documentation | Privacy and Publicity