Collections of Czech libraries contain a vast amount of information. Despite the dominance of textual information, a significant part of our cultural heritage is also information captured graphically, be it drawings, maps, diagrams or other primarily graphic elements.The aim of the project is to open the graphic content of digital libraries to the public. Using machine learning methods, it will be possible to identify the graphic elements contained in digitized documents, to categorize them by type, to add contextual data to facilitate their retrieval and to extend the range of services of our digital libraries by a system for retrieving the graphic elements identified in this way. An important part of the project's outputs will be a tool for finding different images of the same persons and a database of identified persons.