On the influence of low-level visual features in film classification

• DOI: 10.1371/journal.pone.0211406
• Publication Date: 2019-02-22T18:48:57Z
• Authors: Federico Álvarez; Faustino Sánchez; Gustavo Hernández...; David Jiménez; José Manuel Menéndez; Guillermo Cisneros
• Keywords: Databases, Factual; Esthetics; Science; Q; Motion Pictures; R; Video Recording; 02 engineering and technology; Models, Theoretical; Machine Learning; Deep Learning; Sound; Visual Perception; 0202 electrical engineering, electronic engineering, information engineering; Medicine; Cluster Analysis; Humans; Research Article

Background In this paper we present a model of parameters to aesthetically characterize films using multi-disciplinary approach: by combining film theory, visual low-level video descriptors (modeled in order supply aesthetic information) and classification techniques machine deep learning. Methods Four different tests have been developed, each for application, proving the model's usefulness. These applications are: style clustering, prediction production year, genre detection influence on popularity. Results The results are compared against high-level information determine accuracy classify without knowing such previously. main difference with other characterization approaches is that able isolate really understand relevance features and, accordingly propose useful set purpose. This has tested representative number prove it can be used applications.