Leo Impett and Rafael Brundo Uriarte. Seminar Series in Digital Humanities: Methods and Applications of Machine Learning and Computer Vision in Art History

This seminar introduces the principles and practices of two complementary topics: Machine learning and Computer Vision. Machine learning methods are widely used in engineering and natural science fields to model and data, and is increasingly coming into play in the humanities. In the first part of the seminar, I will provide an overview of the ability of systems to learn from past events and how this knowledge can be used in different scenarios. Then, I will discuss some learning methodologies (no technical knowledge required) and the most common solutions in the area. Furthermore, I will explain different applications of these methodologies in the humanities through some case studies.

In the second part, our Guest Speaker Leo Impett will focus on the use of computer vision and image processing in the history of art. He will introduce the prehistory of "visual distant reading" approaches in art history (Gromov & Plyushchev, Marilyn Lavin); outline the remarkable progress computers have made in the understanding of non-photographic images in the last decade; survey current approaches to computer vision in art history (including search engines, automatic metadata generation, computational iconography); and show a number of powerful new technical methods that have yet to find pertinent applications in the humanities (e.g. generative adversarial networks, automatic image captioning). The seminar, which assumes no technical knowledge, seeks to encourage debate on the echoes of twentieth-century formalism in the computational humanities. Highly-professionalised subfields of computer vision (including text recognition and photogrammetric 3D reconstruction), in which the needs of art historians may be better met by private companies than by experimental university projects, will not be discussed.

A Note on the Guest Speaker

Leo Impett is a Digital Humanities Scientist at the Bibliotheca Hertziana. His background is in Computer Vision and so-called "visual computing" (at the Machine Learning Lab, Cambridge; Advanced Technology Lab, Microsoft Research; Image and Visual Representation Lab, EPFL), and especially its intersection with art history and visual studies. In recent years, he has worked with Franco Moretti (Stanford) and Peter Bell (Erlangen) on computer vision for the "distant reading" of the history of art. He is a visiting fellow of Cambridge University Digital Humanities and a visiting scholar at Merton College Oxford.