Emerging AI Applications in Film and News Media
The contemporary film and media landscape is experiencing a disorienting change driven, in part, by new technology. In film, digital effects and the move to streaming content, has led to a revolution in what is possible visually as well as impacting the economics of film distribution. In the news media, there is increasing concern about the proliferation of “fake news” distributed through social media, as well as specific fears about the use of “deep fake” face swapping technology. Computer vision based AI applications have the capacity to identify and analyse video data. This interpretative power is now being brought to bear in the film and news industries, helping to safeguard profits and ensure accuracy in reporting. Supporting this innovation is the diligent training data creation and annotation work carried out by services like Keymakr.
This blog will focus on two specific use cases and show how AI is beginning to have a surprising impact in diverse contexts, as well as how smart data annotation outsourcing can help these efforts going forward.
How AI is transforming product placement in films
Product placement in films and television is a multi billion dollar industry. As the value of traditional advertising placements have shrunk, many brands have instead chosen to invest by making sure their products appear in high profile cultural events that will be viewed by millions. However, product placement can be expensive and prone to interruptions. Films which have their production schedules delayed may have to reshoot scenes so that they feature up to date projects. Similarly, as films are moved to digital libraries and streaming services the products that they feature quickly become out of date and the value of the placement is lost. Applications, powered by machine learning, are being developed in response to this need for flexibility and longevity in product placement. Companies like Ryff, are developing technology that can replace real world objects with virtual, photo-realistic ones in real time. This will allow advertisers to customise and target their product placements, meaning that expensive commitments are now responsive and longlasting.
These developments rely on video annotation. This time consuming process leads to the creation of training data that teaches computer vision models to identify and replace objects across multiple frames in a given scene. Outsourcing this demanding annotation burden could be part of the solution for innovators looking to advance the exciting potential of AI powered product placement.
AI is helping spot deep fakes and fake news
Deep fakes are fabricated, but extremely convincing, images or videos created by AI algorithms. This can mean creating a fake person, showing a person saying something that they did not, or even placing one individuals’ face onto a body in a different video or image. The consequences of the rise of this technology can be pernicious. Individuals can face significant personal embarrassment, or be subject to inaccurate reports and gossip. On a larger scale the presence of deep fakes leads to general distrust of news reporting and can even be used as an excuse for those found to be engaging in real wrongdoing.
In order to combat this AI created problem researchers are turning, ironically, to AI. By being fed huge amounts of training data, including fake and real images, machine learning algorithms are beginning to understand the almost imperceptible differences that separate the false from the true. This important safeguarding work needs to be supported with large volumes of video and image data. Professional annotation providers, like Keymakr, are experts at data collection, and may be best placed to assemble datasets to combat this harmful trend.
Data annotation services help innovators in a complex world
Keymakr utilises proprietary technology, in-house teams of annotators, and multiple layers of quality control to provide data annotation that is precise, affordable and scalable.