University of Cambridge > > Signal Processing and Communications Lab Seminars > On designing robust cost functions – applications to pattern analysis

On designing robust cost functions – applications to pattern analysis

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

If you have a question about this talk, please contact Prof. Ramji Venkataramanan.

Posterior-like functions are widely used in computer vision to infer latent information of interest using data collected from multiple sensors. The problem at hand is originally formulated with a set of equations between the latent and observed variables, and the recorded data is often polluted with unpredictable outliers. In this talk, I will introduce a family of cost functions for the ‘data term’ appearing in a posterior. This family includes very robust models (e.g. Hough Transforms) and the standard likelihood. We will stress the similarities of our framework with other inferential approaches such as L2E divergence. These ideas will be illustrated with applications for 3D reconstruction from multiple view depth images, morphable model fitting, and registration.

BIO: From 1998 to 2001, Rozenn Dahyot (BSc 1996, MSc 1998, PhD 2001 University Louis Pasteur Strasbourg, France) worked for LCPC (French Institute of Science and technology for Transport) as a Research Assistant, developing image analysis techniques to assist in road inspection and management. From 2002 to 2005, she was a Research Fellow in Trinity College Dublin (TCD), Ireland, and University of Cambridge, UK, working on EU funded projects on Multimedia Understanding (e.g. sport video analysis) and Video Restoration. From 2005 to 2008, she was a lecturer in the Computer Science department and since 2008 is a tenure Assistant Professor in Statistics in TCD . Her research currently focuses on robust inference for object detection & recognition, and 3D reconstruction from RGB / depth images. Rozenn is currently the secretary to Irish Pattern Recognition and Classification Society (IPRCS), and its representative in the International Federation of classification Societies. She is the coordinator of the FP7 Industry-Academia Partnerships and Pathways (IAPP) GRAI Search project (2014-18) on the use of Graphics Rendering and Artificial Intelligence for improved mobile Search capabilities.

This talk is part of the Signal Processing and Communications Lab Seminars series.

Tell a friend about this talk:

This talk is included in these lists:

Note that ex-directory lists are not shown.


© 2006-2023, University of Cambridge. Contact Us | Help and Documentation | Privacy and Publicity