HisVis: Scene Detection Pilot Training Set

The De Boer photo collection, provided by the Noord-Hollands Archief, consists of approximately two million negatives and metadata for the period 1945-2004. This training set consists of a subset of 2,545 images from the NHA online repository, an early batch of digitized images, and digitized small negatives.

We took the categorization scheme for Places-365 as a starting point. We combined this scheme with information from the catalog cards that De Boer used. In making these categories, we kept in mind whether there remained a historical and visual consistency in the categories and whether the category would be of use for users of the collection. To gauge the interest of the users, we conducted several interviews.

Together with archivists and cultural historians, we studied the images in the training set and settled on a categorization scheme of 159 categories. During annotation, we noticed the difficulties in distinguishing between scenes that were characterized by a particular object and scenes defined by the location or the action performed in the image.