Comic Onomatopoeia Dataset
Comic onomatopoeia datasets are being developed to advance research in automated comic analysis, focusing on accurately associating text, particularly irregular or fragmented onomatopoeia, with the characters who utter them. Current research employs deep learning models, including scene graph generation, to improve the accuracy of speaker detection and text recognition in comics, often incorporating information about panel order and character context. These datasets and associated research are crucial for enabling advancements in applications such as automated comic translation, audiobook generation, and a deeper understanding of narrative structure in visual media.
Papers
January 30, 2024
June 30, 2023