AI Virtual Museum is an artistic project that, using Artificial Intelligence, proposes to re-materialize the collection of missing paintings that, following the processes of disentailment in the 19th century in Spain, mainly the one known as “Mendizabal Disentailment”, should have been part of the founding funds of the Provincial Museum of Palencia but never became part of it.
The Provincial Museums of Spain collect the most outstanding objects of the Cultural Heritage of each province. Its collections are normally divided into three sections: Archeology, Fine Arts, and Ethnography. However, the Museum of Palencia does not have a Fine Arts section, perhaps because originally, the first endowment consisting of 57 paintings from the province's confiscated monasteries, never became part of the Museum and were lost.
Impacts & Results
The Project consists of three lines of research:
The first one is a kind of detective investigation and comprises the study of all the documentation available on this missing collection of paintings, with two main objectives: first to gather information in order to give visibility and to show the vicissitudes of this heritage in a future exhibition; second, it provides essential information for the parameterization inputs of the AI algorithms.
The clarification of the events that led to the disappearance of the collection, however necessary, is not the main objective of this project. Nevertheless, this documentation phase puts the spotlight on the divulgation of a heritage that was never received by the Museum and therefore is not protected or preserved by the regional government.
A second line of Research, in this case a technical one, is the development and training of the Artificial Intelligence Algorithms that imagine these paintings, re-creating the heritage with the most innovative technology of the 21st century.
The project proposes to investigate and experiment with various aspects of AI and to translate the results of this experimentation into an artistic installation. Specifically, we want to investigate two aspects of AI: on the one hand, studying how machine learning algorithms are capable of recognizing objects [in this case religious paintings] in images and interpreting what they see, and on the other, how this interpretation of what the machines see can be used by these machines, providing an “artificial” point of view, to transform or generate new images, and if this generation can be considered as a first attempt at artificial or computational “creativity”.
That is why the project proposes to reflect on the future of creative and artistic processes and addresses current issues such as the future of Computational Creativity, showing images autonomously generated with AI.
The last but not least of the project activities is the gathering of the artworks images, belonging to some of the main museum collections (supported by a digital identification, description and location system), that are used to train the AI algorithms.
One of the characteristics of some AI algorithms, specifically the Generative Adversarial Networks (GAN), is that they need to learn from a very high number of images. It is necessary that the Artificial Intelligence system that will support this project be able to adequately imagine what the missing artworks could have been like. For this, it is necessary to compile and have a set of images with characteristics consistent with these (period, style, ...) that is made up of thousands of images.