Visual Recognition & 3D

We study core visual understanding — semantic segmentation, detection, and classification with an emphasis on robustness and domain generalization (RobustNet) — alongside 3D perception and neural rendering, including drivable Gaussian head avatars and real-world head swapping.