Movie recommender systems use machine learning to predict what movies a user is likely to enjoy. Several approaches exist, ranging from relying on metadata of the movies to using historical data, from either implicit feedback (for example what movies a user watched) or explicit feedback (for example how users rate movies on a 5-star scale). Different algorithms have been developed to find and use patterns in data to make predictions about what movies a user will like.
Users on the other hand are provided with trailers by movie producers. These trailers allow moviegoers to get an idea of what a film is about, hopefully convincing them to go and watch that movie. The current project aims to investigate how video analysis of trailers can be used to improve recommender system performance. Content based recomendations using such high dimensional data motivates the use of Deep Learning methods.
This project will be supervised by Vlado Menkovski and Mark Graus