Artificial Intelligence has become a buzzword in academia and industry. Some advances made possible by AI have achieved huge publicity – automatic translation, voice controlled assistants or deepfakes, for example. The multimedia research community has seen more accurate image and video understanding made possible, and the TV and media archives industry is not far behind in exploring how these new AI-powered technologies can help them better understand, find, adapt and re-use their large collections of media assets.
Our first workshop on AI4TV, held at the ACM Multimedia 2019 conference and co-organized with the MeMAD project, brought together around 40 participants interested in these topics. From the ReTV project, Johan Oomen (Sound and Vision) gave a keynote on uses of AI technologies in the Dutch media archive. In the oral papers session, Lyndon Nixon (MODUL Technology) presented work on using AI for TV audience prediction and content recommendation; Vasileios Mezaris (CERTH-ITI) presented work on using AI for automatic video summarization.
These are core functionalities of the tools being developed by ReTV: media organisations can know which content will be most successful when published at what time on which vector, then automatically re-purpose that content for that vector into an optimal form for viewer attention. AI certainly forms part of TV’s future and we hope to hold more scientific events on the combination of AI and TV in the future!