Awardee Kathlén Kohn

Press photos

Kathlén Kohn, born on July 2, 1990, in Rostock, Germany. PhD 2018 from the Technical University of Berlin, Germany. Tenure-track, Assistant professor in mathematics at the Royal Institute of Technology (KTH).

Kathlén Kohn simplifies the path from two-dimensional images to 3D with innovative mathematics

The reconstruction of three-dimensional scenes from two-dimensional images is a key technology in modern society, for instance, for building 3D maps, for motion computation for autonomous driving, and when creating special effects for the movie industry. This reconstruction task is a classical problem in the research area of computer vision. The fast development of machine learning has revolutionized the area, and nowadays, deep learning approaches take the main scene in most research conferences about computer vision. This however, also leads to a growing gap between theory and practice.

Kathlén Kohn's research project aims at closing this gap by investigating the underlying geometry of reconstruction problems using methods from algebra. Her interdisciplinary approach achieves foundational and long-lasting results that open new paths for developing faster and more accurate algorithms. Moreover, they strengthen the role of traditional tools that do not rely on machine learning. Traditional tools, predating machine learning, still perform better for some highly geometric 3D reconstruction tasks and have ethical advantages such as better accuracy, reliability, and transparency.

Kathlén Kohn's goal is to identify all reconstruction problems that are efficiently solvable and stable under noise in the image data, i.e., as far as the technology allows. She plans to develop new theoretical and rigorous tools to estimate the intrinsic complexity of those reconstruction problems. Her new tools will be less prone to errors and require less computation time.

A big part of Kathlén’s research focuses on rolling-shutter cameras that make up most of today's consumer cameras, e.g., in smartphones and self-driving vehicles. An upcoming challenge is that fast 3D reconstruction from rolling-shutter cameras where the angle is not known and images are all there is to reconstruct from, is only achieved in restricted scenarios today. Additionally, Kathlén Kohn aims to provide the algebraic and geometric foundations for fast reconstruction algorithms in that setting too.

About Kathlén Kohn

Kathlén Kohn was born 2 July 1990 in Rostock, Germany. She is a tenure-track, Assistant professor at Department of mathematics at the KTH Royal Institute of Technology (KTH). Kathlén started out combining mathematics and computer science at Paderborn University, Germany, and completed bachelor’s and master's degrees in both fields 2009–2015. Kathlén also completed several, paid internships in computer sciences and economic sciences at the company Siemens. In 2014 she attended an exchange student program at Stockholm University. She obtained her Ph.D. in mathematics at the Technical University of Berlin, Germany, for her thesis titled ”Isotropic and Coisotropic Subvarieties of Grassmannians.” in 2018. Subsequently, in 2018–2019 she went on to perform post-doctoral studies, split between the Institute for Computational and Experimental Research in Mathematics (ICERM) at Brown University, Providence, RI, USA, and the University of Oslo, Norway. Kathlén started her current position at KTH in 2019, where she became a docent in mathematics in 2021.

Kathlén Kohn has received several national and international awards and scholarships, e.g. the Best Student Paper Award at the most influential conference, the IEEE/CVF International Conference on Computer Vision 2019, where the top 4 papers were selected from a total of 4.303 submitted papers. Kathlén also received the prestigious Marie Skłodowska-Curie Fellowship in 2019 and The Small Göran Gustafsson Prize for Young Researchers at Uppsala University and KTH Royal Institute of Technology in 2021. Furthermore, Kathlén has received both collaborative funding together with top senior Swedish computer vision researchers, and individual funding from the Knut and Alice Wallenberg Foundation.

Press photos