Unconstrained Text Detection in Manga: a New Dataset and Baseline
This addresses text detection challenges in manga for researchers and applications in comic digitization, though it is incremental with a new dataset and model improvements.
The paper tackles the problem of detecting unconstrained text in Japanese manga by creating a new pixel-level annotated dataset and implementing specialized evaluation metrics. Their deep network model outperforms current text binarization methods in manga across most metrics.
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 binarize text in a comic genre with highly sophisticated text styles: Japanese manga. To overcome the lack of a manga dataset with text annotations at a pixel level, we create our own. To improve the evaluation and search of an optimal model, in addition to standard metrics in binarization, we implement other special metrics. Using these resources, we designed and evaluated a deep network model, outperforming current methods for text binarization in manga in most metrics.