The OCTOPUS project is targeted to broadcasters and content providers in order to help the journalists with out-of-band information regarding the persons who are interviewed.
The main idea of the OCTOPUS project is to avoid “epic failures” that can happen if the journalist does not recognize the person or in case the journalist does not remember exactly the name or the position of the person during the interview process.
For this reason, the FLAME platform is used to assist the on-the-field work of the journalists and the local production unit, for two main aspects.
The first is related to the out-of-band information that is provided in real time regarding persons who could be interviewed. This information is presented directly to the journalist who can now, if desired, access to a complete description of the person in the interview process.
This feature has been achieved thanks to the offloading functionalities of the FLAME architecture that can support many scenarios, mostly based on microservice architectures, in order to locally store the face db as well as run the face recognition algorithm. Thus, OCTOPUS and FLAME demonstrated that heavy and complex tasks can be move on the edge network which also implies a better use of the batteries of the journalist’s terminal.
The second aspect is to support the local production unit, that must select the current flow that should be sent “on-air”. OCTOPUS uses a lot of the offloading functionalities of the FLAME infrastructure even to support the local production unit. Indeed, the face recognition function is running in real-time on all the streams that are produced and then shown to the local production unit which can now also enrich the stream with information on who is interviewed.
The manipulation of the stream by the local production unit, is essential especially for small and medium broadcasters because another side effect is the possibility to manipulate the streams locally and send to the remote production unit only the stream that must be sent on-air with a potential scalability of N:1 (where N is the number of streams produced locally and 1 is the stream sent on-air).
Thus, the required bandwidth for the interconnection of the local production unit and the central office of the broadcaster is much less if compared with classical approaches.