大象传媒

By Michael Armstrong, 大象传媒 Research and Development

大象传媒 R&D’s Intelligent Video Production Tools team has been working with the Springwatch teams using machine learning techniques to help the producers find interesting video clips from wildlife camera feeds. One result of this work is a database of video clips from the programmes which contains the video and a host of metadata, including the outputs from the machine learning. This database is hosted on the cloud, meaning that we can write computer programs to access it and find content and browse it from any location via a web interface. This content has become a valuable resource for creating prototype interactive experiences for us using our Object-Based Media Toolkit.

In the run up to this year’s Springwatch we created a bird identification quiz based on the Autumnwatch camera feeds from the birdfeeders. Because there was a variety of different species visiting the feeders with good identification of each event it was possible to build the quiz using a mixture of automation and manual editing using our StoryFormer tools. The quiz went live on the first day of the programme.

The challenge then was to produce fresh interactive content to accompany the Springwatch programme. The camera feeds for Springwatch are rather more challenging for the machine learning to interpret, as most of them are feeds of birds' nests where only the parents and chicks are visible and the chicks are very different looking from the adult birds. This means that a system trained on images of adult birds has little chance of identifying the youngsters.

So, whilst there was little opportunity for automatically creating a quiz there was still the opportunity to create interactive content. During a conversation with the series editor Joanne Brame and series producer Lucy Bowden they suggested that we could build a quiz around identifying the chicks in the nest. Because it requires a clear view of the chicks it has taken a couple of weeks to find eight suitable clips, using the web interface to the camera feeds. We then reused the structure of our first quiz to manually build 大象传媒 R&D’s first Springwatch quiz: