How to validate Movie Trailers’ success by emotions

In our fast running world we spend lots of time on social sites, and we love to show our feelings and emotions. We can measure the success of a movie trailer based on shares and likes on Youtube. But can we do it in the other way? Can we predict somehow which video trailer will perform better? It is a more complex and more exciting question.

In our talk we share a business case-study how Realeyes can predict which movie trailer will engage the audience more. We use computer vision algorithms for face tracking, machine learning algorithms for extracting emotional and behavioral cues from the tracked faces, and linear regression models for predicting the success of the movie trailers. We developed one universal model for predicting the performance of multiple genre types (comedy, horror, etc). In addition, our novel tool is transparent, easy to understand, and has a low-complexity. Moreover, real movie trailers and real social scores have been used to validate the in-the-wild performance of the model.

Petróczi Attila
R&D and Data Science Project Manager, Realeyes Kft.

Attila is R&D and Data Science Project Manager at Realeyes, where he leads Computer Vision and Predictive Analytics projects. His goal is to bring science from the labs to real life, helps to understand the importance of emotions especially in marketing. Previously Attila worked at Prezi as Analytics Manager and supported the company to be data-driven. As he loves to share his knowledge from Business Analytics and Big Data, he is a guest lecturer of Budapest Business University.