Skip to main content
françaisFR
European Commission logo print header

image Indexing and reTrievAl in the LArge Scale

Description du projet


Search Engines
Video & image Indexing and Retriaval in the Large Scale

Multimedia Archiving

Content providers have growing resources but no adapted tools to make them widely and easily usable for a large public. They are facing problems relative to this impressive growth, and are requiring new ways to efficiently enrich their recourses.

VITALAS is about to unlock the potential of large digitised multimedia databases, and will establish new links between cross-media contents. The project will provide advanced solutions for indexing, searching and accessing large scale digital audio-visual content.

VITALAS is a use-case driven project that will deliver a pre-industrial prototype, allowing intelligent access to multimedia professional archives. Two concrete use-cases will be considered: the first one addressing political/societal news content; the second covering fashion and entertainment news.

The VITALAS technology will not only be applied in business-to-business (B2B) applications, but will also reach out for larger public adoption, by addressing consumers' need for efficient and reliable multimedia content search engines.

Building a professional cross-media search engine

In order to meet the expectations of a broad user community, and provide a robust and scalable multimedia search engine, VITALAS will address three major challenges:

1. Cross-media indexing (automatic annotation) and retrieval
2. Large scale search techniques
3. Visualisation and Context adaptation (personalized services considering both on- and off-line)

This implies the development of efficient and advanced informative content description methods, robust machine learning approaches toward automatic annotation and interactive content search. The methods to be developed will anticipate the on-line user needs (queries) by allowing advanced retrieval of non-previously annotated persons/objects through effective non-textual content description and automatic metadata generation. The scalability issue of searching large scale index feature space will also be addressed, and intelligent visualization techniques will be developed for smart content navigation.

The VITALAS work relies on three core activities:

1. Cross-media indexing and retrieval methods: Efficient cross indexing methods will be developed through semi-automatic multimedia content annotation using several media inputs. VITALAS will put forward appropriate probabilistic retrieval techniques. Advanced hybrid relevance feedback model will be investigated to provide better user target retrieval. VITALAS will consider machine-learning methods together with the development of more informative new content description methods.

2. Interactivity and Context adapting: considering the preliminary use-cases expressed by our content provider partners, approaches which adapt the search space to the user profile and provide interactive functionalities to control the results are expected. Interactive cartographies and video synthetic views should allow users to give feedback, analyse and manipulate the results according to the task being achieved. Off-line user profiles and on-line personalisation will also be used to provide more user satisfaction by expressing users’ subjective preferences.

3. Search scalability issue: to enable search in very large and heterogeneous databases, the system validation will be performed on real and live databases, up to 10,000 hours of television archives and several tens of millions of political/societal news content images.

The VITALAS system technology is not only expected to become a leading B2B application in the field of multimedia content search engines, but it is also foreseen to be derived into a wide array of consumer applications and services tailored for the public willing to access large scale multimedia information. VITALAS will provide advanced solution for indexing, searching and accessing large scale digital audio-visual content. In this regard, the project will confront its technology to real use-cases reflecting the joint concerns of three major European content providers: INA, BELGA. In particular, two concrete use-cases will be considered: the first one addressing political/societal news content; the second covering fashion and entertainment news. To this end, VITALAS will address three major challenges: i) the cross-media indexing (automatic annotation) and retrieval, ii) the large scale search techniques and iii) personalized services considering both on-line (interactive) and off-line context adaptation. This implies the development of efficient and advanced informative content description methods, robust machine learning approaches towards automatic annotation and interactive content search. The methods to be developed will anticipate the on-line user needs (queries) by allowing advanced retrieval of non-previously annotated persons/objects through effective non-textual content description and automatic metadata generation. The scalability issue of searching large scale index feature space will also be addressed, and intelligent visualization techniques will be developed for smart content navigation. As a user-centric system, VITALAS will benefit from a strong involvement of content providers and from external user trials, besides the internal evaluation of individual technological components. Moreover, training of our content provider's archivist partners will be carried out internally, as a first step towards broader vocational training.

Champ scientifique

Programme(s)

Thème(s)

Appel à propositions

FP6-2005-IST-6
Voir d’autres projets de cet appel

Régime de financement

IP - Integrated Project

Coordinateur

GEIE ERCIM
Contribution de l’UE
€ 367 617,00
Adresse
2004, ROUTE DES LUCIOLES SOPHIA ANTIPOLIS -
06410 BIOT
France

Voir sur la carte
Type d’activité
Other
Liens
Coût total
Aucune donnée

Participants (11)

Partager cette page

Télécharger

Dernière mise à jour: 3 Septembre 2022

Mon livret

Permalink: https://cordis.europa.eu/project/id/045389/fr

European Union, 2024