Assoc. Prof. Dr. Bruno Moreschi 

This project refines experiments and implements new methodologies for associating metadata in training images for AI computer vision. With over a decade of experience at the intersection of computer vision and visual arts, including collaborations with various experts, I use the idea of the "dematerialization of art" from the 1960s, where art transcended objects and became a tool to infiltrate and reveal systems such as computer vision.

Recently, my focus has shifted from merely materializing computer vision to transforming it. At institutions like ETH Zurich and Leuphana University, my research moved from speculation to testing new methodologies for training computer vision, such as image tagging.

One innovative approach developed is the Decanonization process, which removes bounding boxes from images. This method provokes reconsideration of computer vision's impact on interpreting scenes. Another concept, "What If…?" tagging, proposes tagging images by considering what is possible rather than probable, introducing a range of potential answers rather than a definitive one.

During the fellowship, I aim to implement these promising methodologies, with exchanges and events to refine and validate the models. This includes systematizing expanded tagging exercises with AI and digital humanities researchers. The ultimate goal is to develop a theoretical framework and draft an article. Additionally, the project envisions showcasing these AI tagging methods in an art exhibition.