Japanese / English


論文の言語 英語
著者 Yuki Daiku, Olivier Augereau, Motoi Iwata, Koichi Kise
論文名 Comic story analysis based on genre classification
書名 Proceedings of 14th IAPR International Conference on Document Analysis and Recognition (ICDAR)
Vol. 03
ページ pp.60-65
ページ数 6 pages
出版社 IEE
年月 2017年11月
要約 Comic readers are attracted to not only pictures of unique characters or beautiful landscape but also deliberated story. Understanding comic story is helpful for a comic retrieval, which allows readers to obtain comics suited to readers’ interest, or creative activities, which demand to generate interesting idea of comic narratives. Therefore, as the method of understanding comic story, we propose a converting method from a real comic story into a novel formatted narrative structure, which uses comic genres as the representation of contents of story. In a converting method, each page in a comic volume is classified into what genre the page portrays by using convolutional neural network. Generally, in machine learning, labeling ground truth on a large number of training samples is necessary, which spends much costs of time and money. In this paper, we propose a learning way, which supports to label ground truth on such many samples. The experimental results show the effectiveness of our proposed converting method.