The 2012 ICSI/Berkeley Video Location Estimation System

TitleThe 2012 ICSI/Berkeley Video Location Estimation System
Publication TypeMiscellaneous
Year of Publication2012
AuthorsChoi, J., Ekambaram V., Friedland G., & Ramchandran K.
Other Numbers3365
Abstract

In this paper, we describe the ICSI/Berkeley video locationestimation system presented at the MediaEval 2012 PlacingTask. We describe the graphical model based frameworkthat poses the geo-tagging problem as inference over thegraph. Our algorithm jointly estimates the geo-locations ofall the test videos, which helps obtain performance improvementsover existing algorithms in the literature that processeach query video independently. Our modeling provides ageneric theoretical framework that can be used to incorporateany other available textual, visual or audio features. The presentation may be found at http://www.icsi.berkeley.edu/pubs/speech/ICSI_MediaEval2012_presentation... .

Acknowledgment

This work was partially supported by funding provided to ICSI through National Science Foundation grants IIS : 1138599 (“EAGER: Collecting Training Videos for Location Estimation with Mechanical Turk”) and CNS : 1065240 ("Understanding and Managing the Impact of Global Inference on Online Privacy"). Additional funding was provided by a KFAS Doctoral Study Abroad Fellowship. Any opinions, findings, and conclusions or recommendations expressed in this material are those of the authors or originators and do not necessarily reflect the views of the National Science Foundation.

URLhttp://www.icsi.berkeley.edu/pubs/speech/ICSI_MediaEval2012_paper.pdf
Bibliographic Notes

Presented at the MediaEval 2012 Workshop, Pisa, Italy

Abbreviated Authors

J. Choi, V. Ekambaram, G. Friedland, and K. Ramchandran

ICSI Research Group

Audio and Multimedia

ICSI Publication Type

Talk or presentation