Overview of Activity and Achievements in 2007

VITALAS is a use-case driven project that aims at providing a pre-industrial prototype dedicated to intelligent access services to multimedia professional archives that will provide the consumer with new technological functionalities. The VITALAS system technology is envisaged to be a B2B tool which will develop technologies that could be applied and re-used for larger public consumer facing multimedia content search engines.

VITALAS will provide advanced solution for indexing, searching and accessing large scale digital audio-visual content (see Table 1 for an overview of the INA archive). In this regard, the project will confront its technology to real use-cases reflecting the joint concerns of three major European content providers: INA, IRT and BELGA.

Table 1: INA's AV archive in 2006 (annual growth: about 35000 hours of television and 16000 hours of radio)

Main achievements in 2007

Market analysis
An in-depth market analysis has been done to assess the market for mixed media search products, to identify which are the key opportunities for VITALAS exploitation, where VITALAS can address market needs and what are its strengths relative to the competition. Based on this analysis we formulated a set of recommendations, which will be taken in consideration in subsequent phases of the project.

Elaboration of VITALAS use cases
As a user-centric system, one of the main objectives of the project was to establish the link between the users’ needs when using advanced technologies in their industrial context and the innovative technologies available in the VITALAS project. Hence the analysis of users’ requirements and expectations, the elaboration of VITALAS use cases and the definition of “Adapted test corpora” were performed during the first six months of VITALAS. A first version of the user interface and of the visualisation modules was developed. The users’ requirements in term of security, access rights and personalization were defined and the architecture of the security module was issued in accordance to the whole system architecture. 

First version of the platform: VITALAS v0
As one of the main objectives of VITALAS is to produce a working pre-industrial prototype system, already during the first semester the definition of the architecture of the VITALAS platform and the specification of the baseline were completed.  A Web service oriented architecture was proposed together with a data reference model for data exchange among services. Furthermore, several business services have been identified and their interfaces have been described. Figure 1 gives a first view of the user interface that is foreseen for the first version of the VITALAS system.

Figure 1 Prototype of the VITALAS User Interface

First year scientific research

• Enabling technologies: specification and development of the V1 SoA modules: content media indexing (audio, video, text), and the new visual local features (interpretable interest points, dissociated local features),
• Cross-media indexing (fusion) with the implementation of a state-of-the-art cross-media indexing tool (audio, text and visual features; combination using SVM), besides literature SOA,
• Cross-media retrieval (Query expressions to represent VITALAS use-cases, initiated search log analysis of Belga image collection, coordination of and participation in INEX multimedia track)
• Implementation of a first version of the required baseline system to allow a first integration of VITALAS components.

User Involvement, Promotion and Awareness

The VITALAS Consortium is pooling together different complementary teams that share common interests ERCIM (Administrative and Financial coordination), INRIA (scientific coordination), EADS, FhG IAIS, CWI, Robotiker, the University of Sunderland, Codeworks, CERTH-ITI, INA, Belga and IRT. Several of these institutes have already collaborated in several past initiatives, such as multimedia understanding through semantic computation and learning, interactive search engine mixing images and text, Digital Media and Assistive Technology, Machine Learning and Indexing, Visualisation and Graphics.

Interested parties in VITALAS, which include audio visual broadcasters, press agencies and researchers in the domain of audio-visual search, are kept up to date on the progress of the VITALAS project through the VITALAS website. This includes the reporting of achievements within VITALAS, dissemination of public deliverables, the presentation of events organized by VITALAS and management information on the composition of the project.

VITALAS project clusters with other related initiatives and networks such as CHORUS or MUSCLE. Members of the Consortium carried out dissemination activities via conferences and events where they presented papers (e.g. ICME International Conference on Multimedia and Expo, CIVR International Conference on Image and Video Retrieval, ECIR European Conference on Information Retrieval).

Future Work

The major activities to come will be to extend integration and collaboration between partners in the VITALAS system in order to improve upon the state of the art, including

• prospection of new methods for generic object recognition,
• visual thesaurus construction,
• video visual features and similarity measures,
• word sense disambiguation,
• speech indexing
• similarity search structures.

The consortium will also seek to make novel contributions in cross-media indexing in particular cross-media representation and cross-media fusion, and will choose the retrieval model whose effectiveness of the ranking component will be measured in international benchmarks.

The next version of VITALAS system will integrate the compliant text, video and audio modules in the platform. The User Interface will be further developed with a more commercially attractive design, and requirements regarding security (data exchange and content protection, user authentication and privacy) will be implemented based on their priority to the overall functionality of the VITALAS system

The first version of VITALAS demonstrator and showcases will be available last semester 2008 and will be implemented in the content provider partners’ premises for trainings and trials.