Friendly faces: Weakly supervised character identification

2Citations
Citations of this article
6Readers
Mendeley users who have this article in their library.
Get full text

Abstract

This paper demonstrates a novel method for automatically discovering and recognising characters in video without any labelled examples or user intervention. Instead weak supervision is obtained via a rough script-to-subtitle alignment. The technique uses pose invariant features, extracted from detected faces and clustered to form groups of co-occurring characters. Results show that with 9 characters, 29% of the closest exemplars are correctly identified, increasing to 50% as additional exemplars are considered.

Cite

CITATION STYLE

APA

Marter, M., Hadfield, S., & Bowden, R. (2015). Friendly faces: Weakly supervised character identification. In Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics) (Vol. 8912, pp. 121–132). Springer Verlag. https://doi.org/10.1007/978-3-319-13737-7_11

Register to see more suggestions

Mendeley helps you to discover research relevant for your work.

Already have an account?

Save time finding and organizing research with Mendeley

Sign up for free