AI4Media – A Centre of Excellence delivering next generation AI Research and Training at the service of Media, Society and Democracy

AI4Media is a center of excellence of leading partners from the fields of AI and media that develops novel technologies and tools for trustworthy AI. Our goal is to address the challenges and risks and, at the same time, to exploit the opportunities of an intensified and widespread use of AI in media, society and politics. This is to ensure that European values for ethical and trustworthy AI are embedded in future AI systems, and establish AI as a foundational technology serving society and media.

Responsibilities of Fraunhofer IDMT

Fraunhofer IDMT’s main contributions lie in the following domains:

  • Development of technologies for audio manipulation detection for disinformation detection and analysis.
  • Development of audio provenance analysis technologiesfor duplicate detection, program analysis, and media research.
  • Development of music analysis technologies to automatically annotate music content for advanced search and recommendation.
  • Development of secure federated learning approaches using homomorphic encryption, which allow participants to train common AI models without having to transfer data to a central location or to reveal sensitive information.
  • Development of privacy-friendly recommender systems that combine usage data analysis and automatic annotation to deliver suitable material for a given context.

Project Partners

The AI4Media consortium consists of 30 leading partners in AI and media (9 universities, 9 research centers, 12 industrial partners) and 35 associate members, forming a network that connects a fragmented European AI landscape in the media sector and promotes long-term cooperation between academia and industry.

Project Lead: Centre for Research and Technology Hellas (CERTH), Griechenland

Research Organizations

  • 3IA Côte d’Azur, France
  • Barcelona Supercomputing Center (BSC), Spain
  • Centre for Research and Technology Hellas (CERTH), Greece
  • Fraunhofer Institute for Digital Media Technology IDMT, Germany
  • Fraunhofer Institute for Intelligent Analysis and Information Systems IAIS, Germany
  • French Alternative Energies and Atomic Energy Commission (CEA), France
  • Idiap Research Institut, Switzerland
  • Institut de Recherche et Coordination Acoustique/Musique (IRCAM), France
  • Joanneum Research, Austria
  • National Research Council (CNR), Italy

Universities

  • KU Leuven, Center for IT & IP Law (CiTiP), Belgium
  • Queen Mary University of London (QMUL), United Kingdom
  • University of Amsterdam (UvA), The Netherlands
  • University of Applied Sciences and Arts of Western Switzerland (HES-SO), Switzerland
  • University of Bucharest(UPB), Romania
  • University of Florence, Italy
  • University of Malta (UM), Malta
  • Aristotle University of Thessaloniki (AUTH), Greece
  • University of Trento (UNITN), Italy

Industry

  • Athens Technology Center (ATC), Greece
  • F6S, Ireland
  • Grassroots Arts and Research (GAR), Germany
  • IBM, Ireland
  • Imagga, Bulgaria
  • InterDigital, France
  • LOBA - Customer Experience Design, Portugal
  • Modl.ai, Denmark

Public Broadcasters and Archives

  • Deutsche Welle, Germany
  • RAI Radiotelevisione Italiana (RAI), Italy
  • Sound and Vision, The Netherlands
  • Vlaamse Radio- en Televisieomroep (VRT), Belgium

Funding

The project is funded under the EU H2020 ICT-48 AI platforms program / call (Grant Agreement No 951911).

Duration

September 2020 – August 2024

This might also be interesting for you

Research topic

Audio and Visual Content Analysis

Extracting meaningful data from audiovisual content

Research topic

Automatic Music Analysis

Audio signal processing and machine learning for music analysis

Research topic

Privacy and Trustworthy AI

Trustworthy media technologies

Research topic

Media Forensics

Trustworthy media content

Audio Forensics Toolbox

A collection of various technologies for the detection of forensic traces, manipulations and synthesis in audio content.