The scientists at Disney Research Center developed a new method that accesses composite audience reactions to movies via facial expression. Named Factorized Variational Autoencoder or FVAE, the method has the potential to reliably predict a viewer’s facial expressions for the rest of the movie after observing an audience within a few minutes.
Zhiwei Deng, a Ph.D. student from Simon Fraser University, said, “The FVAEs were able to learn concepts such as smiling and laughing on their own. What’s more, they were able to show how these facial expressions correlated with humorous scenes.”
FVAEs intimate an explicit ability to predict a viewer’s facial expression for the oddest of the movie after observing an audience member for only a few minutes. Still, the results are exploratory.
Voice Precedent of Disney Research Markus Gross said, “We are all awash in data, so it is critical to find techniques that discover patterns automatically, our research shows that deep learning techniques, which use neural networks and have revolutionized the field of artificial intelligence, are effective at reducing data while capturing its hidden patterns.”
Researchers implement this method for testing in 150 showings of 9 mainstream movies. They chose a 400-seat theater embedded with four infrared cameras to overseas faces of assemblage. Almost 3,179 audience members and 16 million facial landmarks were detected.
Scientist Peter Carr said, “It’s more data than a human is going to look through, that’s where computers come in- to summarize the data without losing important details.”
FVAEs look for audience members exhibiting similar facial expressions throughout the movie. It then learns a set of hackneyed reactions from the entire audience. It seems that audience members display identical facial expressions throughout the movie. After that, it learns from the hackneyed responses of the whole audience.
It also explains how audience members should react to a given movie based on strong interdependency in revulsion between audience members.
Scientist Stephan Mandt said, “FVAEs to predict a viewer’s facial expression for an entire movie based on only a few minutes of observations”
The pattern recognition technique inside it is not only limited to faces but is also used on any time series data collected from a group of objects.
Yisong Yue, an assistant professor of computing and mathematical sciences from the California Institute of Technology, said, “Once a model is learned, we can generate artificial data that looks realistic. For instance, if FVAEs were used to analyze a forest- noting differences in how trees respond to wind based on their type and size as well as wind speed -those models could be used to simulate a forest in animation.”
