The popularity of artificial intelligence is rapidly growing as more and more companies invest in data collection and analysis, and Disney is no exception. A project funded by Disney Research introduced an AI system that utilizes facial recognition to predict how audiences will react to movies.
Through the use of factorized variational autoencoders (FVAEs), the facial tracking system will be able to predict an individual’s facial reactions for the entire length of a movie, based on just 10 minutes of observation and analysis. The researchers involved in the project built a dataset of over 16 million faces by placing 4 infrared cameras at 150 showings of 9 mainstream Disney movies. Though the AI system is relatively new, it has already performed better than other predictive analysis techniques when tested on several new audiences.
As a result of such technological advancements, AI has the ability to understand human behavior and reactions better than human market researchers, who have been attempting to track this data for several years. “Understanding human behavior is fundamental to developing AI systems that exhibit greater behavioral and social intelligence,” said Caltech’s Yisong Yue, one of the primary researchers on the project. The AI facial tracking mechanism acts as a recommendation system, a type of collaborative filtering not far from Amazon’s method of suggesting products to its consumers.
The data could have huge implications once it is integrated into movie making. Hollywood executives can integrate the system into production and editing processes, essentially “hacking” films to ensure that movie scenes will achieve the desired emotional reactions from audiences.