Genre Classification of Movie Trailers using 3D Convolutional Neural Networks

There is a constant influx of movies every year in the entertainment industry. A crucial ingredient that helps movies succeed at the box office is its trailer. A trailer helps a viewer form an opinion about the movie and is often an accurate representation of the movie. Thus, genre identification of movie trailers is a significant problem that can be solved for movie categorization and censorship. In this paper, we propose using a 3D CNN for solving this problem. The advantage of 3D CNN is its ability to capture the spatial as well as the temporal information present in a movie trailer. We achieve an accuracy of 82.14% which is greater than other models that utilize just 2D CNNs.