Keynote 1 - Scaling 3D: Understanding Model, Data, and Loss
Abstract
This talk will take stock of where we stand in 3D computer vision along three key axes: data, model architecture, and supervision signals. Data is perhaps the most obvious scaling dimension, but also one of the hardest to obtain in sufficient quantity and quality. At the same time, 3D data comes in many forms, including point clouds, meshes, and depth maps, each posing distinct challenges for representation learning and model design. Finally, large-scale vision models such as DINO have achieved impressive performance through self-supervision, whereas most 3D models still rely heavily on direct supervision, which is often expensive and difficult to scale. The talk will summarise findings from our recent and ongoing work, while also drawing broader connections across the field.
Bio
Christian Rupprecht is an Associate Professor of Computer Science at the University of Oxford and one of the leads of the Visual Geometry Group (VGG). His research focuses on computer vision and machine learning, particularly 3D reconstruction, generative models, and visual understanding. He received his PhD in 2018 from the Technical University of Munich. His work received multiple awards, including two Best Paper Awards at CVPR. Prof. Rupprecht is also Chief Research Officer at Third Dimension AI, a company building spatial generation technologies at the intersection of 3D and world modelling.