ITU researchers receive Villum grant to improve image recognition systems

Professor Sebastian Risi and Associate Professor Anders Sundnes Løvlie have just received a Villum Synergy grant for a project to improve image recognition systems by training on nonphotographic visual art.

Anders Sundnes LøvlieSebastian RisiDigital Design DepartmentResearchgrants

Written 23 September, 2021 08:02 by Jari Kickbusch

Image recognition systems are highly specialised and perform poorly when confronted with image types different from the data used in training – in particular when dealing with nonphotographic images. In a recently funded project, Professor Sebastian Risi and Associate Professor Anders Sundnes Løvlie, both from Digital Design Department at the IT University of Copenhagen, will address the issue known as the cross-depiction problem:

- Visual works of art show the human capacity for recognising and interpreting images, from a realistic photo to an abstract painting. We hypothesize that developing visual recognition algorithms that can recognise objects in works of art may give us a deeper understanding of computer vision compared to human vision. It may ultimately help us develop more general and robust vision algorithms, says Sebastian Risi.

The project
The two researchers have just received a Villum Synergy grant of DKK 3 million
to realize the project Algorithmic Ways of Seeing: Improving Image Recognition by Training on Art Images. During the project period, the two researchers will address two important obstacles in cross-depiction object detection: Firstly, the lack of sufficiently large datasets of nonphotographic images for training. Secondly, the ethical and cultural differences exposed when comparing computer and human reception and interpretation of images.

- Working in this field, we will collaborate with people who are experts in interpreting visual art: artists, curators, and the art audience. We hope that this collaboration will be of value to the field of art studies as well, both in terms of technology, which provides us with new ways to experience art, as well as knowledge about the human relationship to both possibilities and limitations of Image Recognition technology, says Anders Sundnes Løvlie.

About Villum Synergy
Villum Synergy is Denmark’s largest programme for data-driven interdisciplinary research. In 2021, 12 projects will receive DKK 45 million in grants through the programme. Learn more on Villum Fonden’s webpage.