Aiding Provenance Research. A Computer-Assisted Image Retrieval in Auction Catalogs

Provenance research examines the origin of objects and aims to reconstruct their ownership history. For this purpose, researchers increasingly use online resources such as the database German Sales, which contains auction and sales catalogs, through a full text search. While this is very valuable, a text-based search has limitations due to missing or varying information. This motivated us to search for a different, image-based method which enables searches for identical and similar images in large data collections using neural networks. We provide qualitative and quantitative results showing the effectiveness of our method and discuss the technical conceptualization of image similarity underlying the presented algorithm.