Unconstrained Text Detection in Manga

TL;DR

This paper presents a new deep learning approach for pixel-level text detection in Japanese manga, addressing unique style challenges and introducing a specialized dataset and evaluation metrics, leading to improved detection performance.

Contribution

The work introduces a novel dataset with pixel-level annotations for manga text and develops a deep network that outperforms existing methods in this domain.

Findings

Deep network achieves superior detection accuracy in manga.

Custom metrics effectively evaluate pixel-level text detection.

Created the first annotated manga dataset for text detection.

Abstract

The detection and recognition of unconstrained text is an open problem in research. Text in comic books has unusual styles that raise many challenges for text detection. This work aims to identify text characters at a pixel level in a comic genre with highly sophisticated text styles: Japanese manga. To overcome the lack of a manga dataset with individual character level annotations, we create our own. Most of the literature in text detection use bounding box metrics, which are unsuitable for pixel-level evaluation. Thus, we implemented special metrics to evaluate performance. Using these resources, we designed and evaluated a deep network model, outperforming current methods for text detection in manga in most metrics.