Multimodal film music dataset

This dataset was initially developed under Claire Arthur’s supervision at Georgia Tech University. It comprises a curated collection of films from 12 countries, 5 genres, and 3 decades. We use machine-learning techniques to analyse different aspects of the music, visuals, and text, and see how they relate.

In his PhD, Farshad is substantially extending this dataset to include many more computationally extracted features, covering diverse topics such as harmony, narrative structure, and camera movement.

This app is temporarily password protected while the corresponding paper is under review. For the password, please contact Farshad directly (fj303@cam.ac.uk).