Shaping the future of Media with AI and Data

The Data Technology Seminar DTS 2025 of the European Broadcasting Union (EBU) will take place from March 11 to 13, 2025, in Geneva. It provides a platform for experts from academia, industry, and broadcasting to exchange ideas on the topics of AI, data, and media technology.

Fraunhofer IDMT will give a presentation on AI solutions for media analysis, and Hanna Lukashevich, head of the Semantic Music Technologies group, will also support the program committee of DTS 2025.

The automatic analysis of audiovisual content is a central focus of numerous research projects, products, and services. The presented solution integrates video and audio analysis methods to automatically enrich large volumes of A/V content with metadata, specifically emphasizing individuals in the media. How often do politicians appear on various channels, and what statements do they make? How equally are all genders represented in specific programs? How can I locate all programs featuring a particular person in my archive? While large media archives contain vast amounts of data, powerful automated analysis tools are essential for efficient access and usability.

The technology presented offers in-depth insights into individuals' media presence by combining facial recognition and speaker recognition to capture and evaluate their appearances in video and audio independently. This reliable identification facilitates the generation of detailed reports on screen presence, speaking shares, and diversity metrics based on perceived gender. In addition to analyzing thousands of hours of television content, preparing and presenting extensive results introduces further challenges. An interactive dashboard is being developed to display these analysis results, enabling various visualizations and rapid data filtering. The presentation will cover fundamental challenges in cross-modal analysis of large A/V datasets, along with the system architecture, process chain, and selected components of Fraunhofer IDMT’s cross-modal content analysis solutions.

March 11, 2025, 3:20 p.m.