Computer vision

Research group leader: Professor Fredrik Kahl

Our researchers are listed below.

About the research area Computer vision

The aim of computer vision and image analysis is to make computers understand images, videos and other image-like data. Given that the human vision system is one of the main sensory organs, enabling computers to see and understand images has a wide range of applications; example applications include interpreting a traffic scene in order to safely drive a vehicle, recognizing an object and its pose to be grasped by a robot arm, and localizing a tumor in a 3D magnetic resonance image. Machine learning has enabled some recent major breakthroughs in computer vision, which have pushed the research frontier and led the way to human-like performance for many applications. Still, there are many research challenges ahead that need to be addressed.

The problems arising in computer vision are complex and require a wide spectrum of tools from geometry, statistics and numerical methods. To a large extent, the research in our group is concerned with developing new methods and theory. It can be the analysis of an optimization problem that is central in machine learning or the development of suitable mathematical models. There are also projects related to industrial applications. We work closely with other research groups nationally and internationally, within our field and with domain experts in medicine, robotics and mathematics. We also work with industry, ranging from involvement in start-ups and to research collaborations with large companies.

Research area 3D Computer Vision

Reconstructing three-dimensional geometrical models of the world is a classical problem in computer vision. We develop algorithms for Structure-from-Motion of rigid and dynamic scenes, multi-view stereo and visual localization.

Research area Machine Learning

We are interested in developing machine learning techniques, in particular, based on deep learning for semantic segmentation, recognition and scene understanding. We are also interested in geometric deep learning, that is, problems involving learning techniques for point clouds, graphs, meshes and other geometric primitives.

Research area Mathematical Models and Optimization

We study and develop mathematical models and corresponding optimization algorithms arising from applications. Examples include bundle adjustment, robust estimation and outlier handling, rank and sparsity priors, and convex relaxations.

Research area Medical Image Analysis

In collaboration with medical partners, we develop methods for anatomy segmentation, robust registration and computer-aided diagnostics. We work with several different types of medical images, including CT, MRI, PET, ultrasound and microscopy, and target medical applications, for instance, coronary artery segmentation and image-guided surgery.

Sort by


Page manager Published: Mon 15 Mar 2021.