Machine Tagging for Personal Multimedia Contents

In a future digital home where a media-center-type device is expected to aggregate large amounts of personal multimedia content such as pictures or home movies, it is desirable to have some tags about the media content available to the consumer to allow them to easily find a picture or movie. These tags must be machine-generated, if at all possible, and must convey meaningful information about the media. In other words, the media content is translated to a subjective or perceptual form of data to aid in the retrieval by a human.

In this project we propose a tagging and retrieval system for personal photos. Each photo is automatically tagged with significant information. “Indoor” is used, for example, if the photo was taken indoors. This should allow users to conduct a very rapid and intuitive text-based search.

A web-based GUI:
http://thor.ece.ucsb.edu/php/photo/login.php
(The development is still ongoing. Mozilla Firefox is recommended for better visualization.)