News & Updates

Gave a keynote talk "3D Vision: Done & Dust3d?" at the International Conference on 3D Vision (3DV 2026) in Vancouver.

Beyond Benchmarks of IUGC: Rethinking Requirements of Deep Learning Method for Intrapartum Ultrasound Biometry from Fetal Ultrasound Videos

Papers accepted: VGGT-Omega, Particulate, Scene-Centric Unsupervised Video Panoptic Segmentation, LagerNVS, LitePT, and SceneScribe-1M

Papers accepted: What Happens Next? and S3OD

Awarded an ERC Starting Grant for the project 'Volute — Visual Omniversal Learning from Universal Teachers'. The grant is worth up to €1.5 million over five years and addresses the shortage of labelled training data for large computer vision models.

Unlocking In-Context Learning for Natural Datasets Beyond Language Modelling

Papers accepted: DIY-Corr, SEVA, AnimalClue, PuppetMaster, DSO, SSC, CoTracker3, CNS-Bench, and SynCity

Our paper VGGT won the best paper award at CVPR'25 out of 13000 submissions.

Papers accepted: AnyCam, DualPM, FLARE, Unbiasing Video Datasets, CUPS, and VGGT