Research in the GAPinDNNs group has led to companies that bring geometric deep learning and neural rendering into industry and healthcare. These ventures are founded by members of the group and keep close ties to our research.
Expert feedback on every ultrasound scan
Medical imaging · AI for ultrasound training
Kymeon develops an online platform for point-of-care ultrasound training: clinicians upload their scans, experienced instructors review them, and every review counts toward certification. Building on the group’s work on geometric deep learning, Kymeon is researching AI models that give learners live feedback on image quality while they practise eFAST ultrasound in trauma care.
Interactive, editable virtual worlds from raw sensor data
Neural rendering · Simulation for autonomous driving
Recohere automates the creation of high-fidelity digital twins from raw sensor data, using neural rendering techniques such as Neural Radiance Fields (NeRFs) and 3D Gaussian Splatting. The resulting virtual worlds are interactive and editable, and are used for simulation and testing of autonomous driving systems.