At the IEEE/CVF Conference on Computer Vision and Pattern Recognition (CVPR) 2024, the paper “SpiderMatch: 3D Shape Matching with Global Optimality and Geometric Consistency” by authors Paul Rötzer and Prof. Dr Florian Bernard was awarded ‘Best Student Paper Runner-Up’. This makes their work one of the ten most outstanding of over 11,500 papers submitted to the conference. Prof. Dr. Bernard heads the ‘Learning and Optimisation for Visual Computing’ working group in the Visual Computing department at the Institute of Computer Science; Rötzer is a doctoral student in his working group.
The CVPR conference took place from 17 to 21 June 2024 at the Seattle Convention Center in Seattle, Washington State. It brings together experts from science, business and industry to discuss the latest developments and advances in keynote speeches, lectures, poster sessions and numerous workshops. The conference's focus on high-quality research and innovation makes it an important event for anyone working in the field of computer vision and pattern recognition. According to Google Scholar Rankings, it is one of the two most important publication venues for all specialist disciplines and the number one conference in the field of ‘Engineering & Computer Science’. An award at this conference thus emphasises the high quality and progressive nature of Rötzer and Bernard's research work.
The award-winning paper ‘SpiderMatch’ presents a novel approach to 3D shape matching. A fundamental problem in 3D shape matching is to find matches between surfaces of 3D objects - a task with applications in medical imaging, graphics and computer vision. The most important innovation of Rötzer and Bernard's work is that geometric consistency is preserved. A property that was neglected in previous methods due to their complexity. “Imagine two organs, like the liver, heart or lungs, and you match them from different people,” explains Prof. Dr. Bernard in the CVPR event magazine. “You take the shapes and want to train a statistical shape model. If you didn’t have this
geometric consistency, the deformation from one to the other would lead to self-intersecting sections which aren’t anatomically plausible." Their new approach in 3D shape matching is based on an optimization problem of finding the shortest path. With their solution, the authors can show that they can solve this efficiently up to global optimality.
In addition to honouring the best research work, CVPR 2024 also offers a series of lectures and panels with leading experts from science and industry. This year's topics include the social challenges of AI, artificial biodiversity and how to drive customer innovation and industrial acceptance of computer vision. CVPR 2024 is supported by a number of high-calibre sponsors, including leading technology companies such as Google, Microsoft, Facebook (Meta), Sony, Qualcomm, Intel, Apple, and Amazon.