2009 Theme

The theme for ODCSSS 2009 "Technologies for bridging the digital-physical divide: sensing the environment". Each ODCSSS student is engaged in a research project with a faculty member and mentor which provides them an opportunity to experience research. The selection of interns is highly competitive but we encourage anyone interested in research who is eligible to apply.

Interactive Content-Based Video Search

Odysseus: 
2010
 
Background:
When we search on the internet, using a search service such as google, we are conducting a content based search.  That is, the words we use for a query are matched against the words which are contained in the webpage we are trying to locate.  Current video search engines, such as YouTube, rely on metadata to enable search against video, very similar to how we used to locate books in a library catalogue.  This project's aim is to construct a content-based video search engine.  That means that this search engine will match user's queries against the actual signal of the video.  Because of this, the user is no longer restricted to using just text to search for video, they can use images or other segments of video as queries to find the content they are looking for.  A popular illustration of this on the web can be found at:  http://www.TinEye.com to provide a working example of the sort of systems we build.

Aim:
Our aim is to construct an interactive video search engine, as part of the TRECVid evaluation campaign, organized by NIST in the USA.  TRECVid attracts research groups from around the world, both academic and industrial, and is the premier forum for the development of multimedia search engines.  The search engine we will build will be to answer a 'known-item' query.  That is, the system will have a user, and that user will be required to find the one exact matching video for a particular query.  The development of the search engine will involve a combination of many discrete processing components.  The aim of the successful candidate will be to assist in the development of various components, and depending upon the timeline, assist in the successful running of a valid, scientifically sound, user experiment.

Outcomes:
The candidate upon completion will have had exposure to a collaborative research environment and to concepts which will be of use later in the candidates career.  Ideally the candidate will have gained knowledge on experimental methodologies, which will be of assistance when later in their careers, either in industry or academia, the candidate is required to demonstrate the effectiveness of something they have developed.

Requirements:
The successful candidate will have a core competency in Computer Science, ideally with experience in Java.  The candidate will be required to operate as part of a team, to take the initiative when difficulties are encountered and seek help from fellow team members.  

Further Reading:
As part of their application, candidates are expected to have familiarized themselves with the following reading materials, which will provide a more detailed explanation of TRECVid, its goals and its operation:

TRECVid call for participation:  http://www-nlpir.nist.gov/projects/tv2010/tv2010.html
Evaluation campaigns and TRECVid:  http://doras.dcu.ie/415/
An overview of what makes a video search engine: http://doras.dcu.ie/376/
Supervisors and Mentors: 
Prof. Alan F. Smeaton
Dr. Peter Wilkins
Host: 
DCU