Patent Image Databases

The first dataset contains 2000 patent images extracted from patent documents provided by the European Patent Office (EPO). This dataset was manually classified into different categories in order to perform evaluation experiments in the context of the EU project PATExpert.

Download the first patent image database here. The settings of the PATExpert experiments are available here.

The second dataset includes 1042 patent images extracted from arround 300 patents from the European Patent Office (EPO) and the United States Patent and Trademark Office (USPTO). This dataset was mannually annotated with 8 concepts.

Download the second patent image database here.