Disney want to know how you enjoy their films, at IEEE’s Computer Vision and Pattern Recognition, Disney Research and Caltech highlighted the way they track facial expressions of people watching films.
The tech is so advanced they claim that they can predict the face’s future expressions through the remainder of a film, its called “factorised variational autoencoders” (FVAEs)
To make this possible, researchers used infared cameras to the audiences of 150 showings of 9 films, including a lot of Disney movies! Like Zootopia, Big Hero 6 and the recent Star Wars films.
After compiling this huge amount of data, they went back and tested it on some more moviegoers, after an initial set up period the FVAEs were able to predict when an audience member (remember these are brand new fresh audiences) was about to laugh or smile as tied to significant moments in each of the films. They hope to predict other emotive moments in films such as fear and sadness. How this data will be used is unclear, but it could help them better understand audience reactions than human based market research.