Comic Story Analysis Based on Genre Classification
Yuki Daiku, Olivier Augereau, Motoi Iwata, Koichi Kise · 2017
People who are reading comics are interested not only by the estheticism of the graphics but also by interesting stories. Therefore, understanding the detail of comic story is an important step to build comic retrieval systems based on readers' interest. As a method of understanding comic story, we propose to convert comic story into a novel formatted narrative structure, which uses comic genres as the representation of the contents of story. Each page of a comic volume is classified into a genre by using convolutional neural network. Generally, in machine learning, labeling ground truth on a large number of training samples is necessary, which costs time and money. In this paper, we propose a story analysis method which can describe a comic story as the sequence of genres with relatively low manual cost. The experimental results show the effectiveness of our proposed method.