Abstract
This paper introduces the Mainz Cuneiform Benchmark Dataset (MaiCuBeDa) and the Mainz Cuneiform Benchmark Dataset for the Haft Tappeh Collection (MaiCuBeDa HT), two datasets of cuneiform sign annotations on renderings of 3D models of cuneiform tablets. The first dataset includes annotations of the Frau Professor Hilprecht Collection, a collection of cuneiform tablets from different time periods. The second dataset comprises the first publications of cuneiform sign annotations from the Middle Elamite period, as excavated at the city of Haft Tappeh, Iran. Both datasets are prepared for use in machine learning tasks, such as cuneiform sign recognition, and, due to their rich metadata, also for, e.g., time period classifications of cuneiform tablets. At the same time, they might be used for paleographic studies in Assyriology of their specific time periods. All annotations are also available through a web interface, allowing experts to suggest corrections for further dataset iterations.