We are interested in a broad range of topics at the interface of visual computing, mathematical optimisation and machine learning:
3D Shape Analysis and Geometry Processing
Geometric data is a crucial part of many tasks in computer vision and computer graphics. Our interests include shape deformation models, shape editing and manipulation, correspondences, geometric data curation, and many more.
Correspondence Problems
We develop optimisation and deep learning approaches that are able to identify corresponding parts between objects (e.g. for images, graphs, surfaces, etc.).
Graph-based Modelling & Graph Algorithms
Many visual computing problems can be formalised in terms of graph-theoretic problems, so that they can be solved using graph algorithms. For example, numerous variants of matching problems can be addressed by finding shortest paths in product graphs.
Deep Learning
We study how to incorporate human knowledge (e.g. in the form of 3D geometry, or physics) into deep learning systems, so that they can focus on learning the unknown, rather than wasting capacity on rediscovering what humans already know.
Large Data Collections
We study how we can adequately analyse and process large data collections (e.g. images, shapes, point clouds, matrices, etc.). This includes efficient algorithms for multi-graph matching, multi-shape matching, and the synchronisation of large matrices (e.g. transformations or permutations).
Applications
While our primary focus lies on foundational research, many problems that we study are driven by real-world applications, including (bio)medical image analysis, motion capture, AR/VR, or 3D reconstruction.
