Automatic Speech Balloon Detection and Segmentation for Comic Books

You might find that reading comic books is complex and confusing, if you are a machine. If both of those ifs afflict you, you might seek relief by reading this new study:

Deep CNN-based Speech Balloon Detection and Segmentation for Comic Books,” David Dubray, Jochen Laubrock, arXiv:1902.08137v1, 2019. (Thanks to Mason Porter for bringing this to our attention.) The authors, at the University of Potsdam, Germany, explain:

“We develop a method for the automated detection and segmentation of speech balloons in comic books, including their carrier and tails. Our method is based on a deep convolutional neural network that was trained on annotated pages of the Graphic Narrative Corpus…. Qualitative results suggest that wiggly tails, curved corners, and even illusory contours do not pose a major problem. Furthermore, the model has learned to distinguish speech balloons from captions.”

Here’s further detail from the study: