Disney Uses AI For Animated Character Face Detection

JAKARTA - So many technological innovations have been created from machine learning or what is often known as artificial intelligence (AI), one of which is a face detection tool.

However, this does not only work in humans, it is said that the Disney animated film production house will use this technology to recognize the faces of their animated characters.

Quoting from Endgadget, Sunday, July 19, the Disney Direct-to-Consumer & International Organization (DTCI) team said this AI development was used to make it easier for them to create digital archives for the millions of films and characters that Disney had created. This system is called the Content Genome.

Content Genome is similar to the profile search feature of a character on Google. Later this Content Genome will contain content metadata that can help animation creators to map animated characters in Disney archives.

"So when an animator wants to find a character reference it can be helped by this platform. So, animators will be able to easily find the metadata needed," said Director of Research and Development at DTCI, Anthony Accardo.

According to Accardo, this project has been started since 2016. At that time, Disney animators took hours to find references and metadata for every character Disney had worked on.

Even so, Accardo admits that building a system like this is not an easy thing. He and his team had to compile a very precise material classification of every character metadata in the Disney archives.

"Especially if you're going to generate lots of different metadata for lots of different attributes. You have to start thinking about how you are going to manage those requirements and labels," explains Accardo.

With the help of Conten Genome, AI technology will immediately classify each animated character based on their tagging pipeline or labeling line. Where the animator can identify the metadata of each character based on the storyline and certain characteristics.

"Marking is also the only way to identify lots of stories and highly contextual character information from structured data, such as storylines, character archetypes, or motivation," he added.

In the future, Accardo and the team hope to further expand the capabilities of this software module to understand the characteristics of each work and animation created. But before that happens, of course, more research and development is needed.

"This is certainly AI machine learning in the context of understanding the characteristics of the image. Of course it is difficult, but you have to start with things that can be identified clearly," concluded Accardo.