Fox Film Studio developed an AI system that can tell you whether a movie will be hit or flop. Movies are at some point is like gambling, you don’t know at the start whether your next 2 hours are will be of fun or torture.
Implementation of such a Machine Learning system not only lots of money but also the time of the audience. It is an algorithmic method capable to predict what kind of audience will go to see a movie. The cool part of this AI is that it does all this thing things just by analyzing a movie trailer.
As we all know that trailers are very important for movie success. For us, it is a short preview of any movie. It tells us about the storyline and what actors are there in the movie. As a marketing point of view, trailers are the most important tool. It ultimately generates movie awareness.
So, if you’re trailers is working as a fire in the audience, chances are your movie is next blockbuster.
Now, let’s talk about the Fox Film Studio. The studio wants to make trailers which must be loved by the audience. And as necessity is the source of invention, they built a machine learning method called Merlin Video.
Merlin Video method predicts what audiences will most likely watch movies with the help of its trailer. It analysis the features of the trailer such as illumination, color, faces, objects, and landscapes. After analyzing all this feature Merlin generates a dense representation of a trailer.
With the help of attendance and demographic data, Merlin accurately forecasts the audience attendance for a movie. Merlin Video has a logistic regression layer that merges the distance-based Collaborative Filtering model to generate the probability of audience attendance.
Researchers trained Merlin’s convolutional neural network over hundreds of movie trailers released in the past along with records, consist attendance counted for each movie.
By feeding appropriate representations to the neural networks, it starts to discover significant links between the features of the movie trailer and future audience preferences.
Merlin Video method uses NVIDIA Tesla P100 GPUs on the Google Cloud, with the cuDNN-accelerated TensorFlow deep learning framework.
Merlin has shown pretty interesting results, but it is still in the development stage. Merlin is not able to capture temporal info.
The Fox Film Studio researchers are focusing on developing a system that uses both text and video features to predict a movie’s success.
Fox Film Studio work is published on ArXiv and was authored by Miguel Angel Campo-Rembado, SVP of Data Science & Analytics at Twentieth Century Fox Film Corp, and his team.
You can read the paper from here.
Read also, you may like it :