What Can We Learn From Movie Colors?

Abstract

We applied supervised learning methods to classify movies into one of four genres based solely on colors present in the film. Data was obtained from screen captures of the Netflix preview window at 10 second intervals. Our final feature was X = {512-color histogram, avg. color change}. Our genre classes were Y = {Action, Animation, Horror, Romance}. We attempted Multinomial Naive Bayes and 1-v-all SVM to predict the genres of a film based on its color features. We also attempted Unsupervised Learning K-Means Clustering algorithm on our movies to observe patterns across movies.

Extracted Key Phrases

12 Figures and Tables

Cite this paper

@inproceedings{Chan2015WhatCW, title={What Can We Learn From Movie Colors?}, author={E. Chan and John Lee and Rajarshi Roy}, year={2015} }