We introduce the first benchmark dataset for slide-page segmentation. Presentation slides
are one of the most prominent document types used to exchange ideas across the web,
educational institutes and businesses. This document format is marked with a complex
layout which contains a rich variety of graphical (eg diagram, logo), textual (eg heading,
affiliation) and structural components (eg enumeration, legend). This vast and popular
knowledge source is still unattainable by modern machine learning technique due to lack of …

In Figure 1, we show the average surface size of each class. Namely, we pick the set of
images that contain at least one pixel of each class separately. From these we average the
percentage that each class covered the image. We see a high variance of region sizes
between classes. For example, for affiliation and the current date we have a rather small
region, while for most of the image classes we have a high number of pixels, as some
images almost capture the whole page. As we see, the class 'maps' has the largest average …