Traditionally, television was produced entirely by professionals, with little audience involvement. However, since the public can now easily create and share videos, television broadcasts increasingly include public-created content. Given the march towards larger and more vivid screens, combined with a growing trend of on-demand viewing, public-supplied video generally does not meet broadcast-quality requirements. The EU-funded COGNITUS project developed software to automatically convert user-supplied videos to an ultra-high definition (UHD) format suitable for broadcast. Ordinary HD video consists of 1 920 columns by 1 080 rows of pixels. At 3 840 by 2 160 pixels, UHD resolution is 4 times that resolution, and almost equal to the highest cinema quality. Fixing any video Project partners developed technologies making this conversion possible. The tools include mobile device apps for public contributors, a web-based app for producers, including smart content discovery and plot-authoring tools, plus several other modules. The team also demonstrated the ease with which converted video could be woven into broadcasts of large events, and the value for broadcasters of doing so. The software essentially makes amateur footage look and sound more professional. “Amateur video can include camera shake, low resolution, and sound distortions such as wind noise,” says Dr Marta Mrak, project coordinator for COGNITUS, “all of which our software can automatically correct.” The tools also improve metadata, frame rate, dynamic range adaptation, audio and file compression. Trials of applied intelligence Given the complexity of user-supplied formats, and the number of potential problems, manual conversions are not always possible. The new COGNITUS software therefore includes AI algorithms to automate the quality checks and enhancements. Requests for public videos can also overwhelm the broadcaster. “So, we developed tools to help producers quickly find videos relevant to the programme they are making,” explains Dr Mrak. “The tools include automated ways of enriching video metadata to make it suitable for smart searches.” The searches may include text terms plus visual or audio similarities. COGNITUS trialled the system at public cultural events in the United Kingdom – including London Pride, the Edinburgh Festival Fringe and the Royal Edinburgh Military Tattoo – and upon various different football leagues in Greece. In all cases, members of the public used the COGNITUS mobile app to upload video they had shot. This provided a personal viewer perspective of spontaneous events professional camera crews could not capture, and at a quality satisfying television producers. The project successfully demonstrated the platform’s performance under real conditions, in many respects exceeding expectations. The European Commission’s Innovation Radar for key innovations commended many of the software’s modules. Research and commercial collaborations The team also successfully collaborated with other Horizon 2020 projects. One of these was InVID, which helps television news producers authenticate user-supplied news content. COGNITUS partners are keen to exploit the project’s outcomes commercially. Researchers are currently engaging with decision-makers to bring the system to market. The project-developed tools offer a way for the public to become involved in the production of broadcast content. Including amateur video of public events makes the audience feel better represented and more engaged with the broadcast. By enabling such public inclusion, COGNITUS also helps lower the costs of television production.
COGNITUS, user generated video, television, UHD, machine learning, automated, broadcast content, artificial intelligence