Cordelia Schmid

Papers

Total Citations

H-Index

About

Cordelia Schmid is a distinguished computer vision and robotics researcher whose work spans visual representation learning, robot manipulation, and the transfer of human knowledge to robotic systems. Her recent contributions reflect a forward-looking focus on bridging the gap between 2D and 3D understanding in machine perception. In her 2024 work **SUGAR**, Schmid and her collaborators tackled a critical limitation in robotics pre-training — the over-reliance on 2D visual representations — by developing methods to learn generalizable 3D visual representations from Internet-scale data, enabling robots to better handle occlusions and precise object localization in complex scenes. Complementing this, her work on **ViViDex** explores how multi-fingered robotic hands can acquire dexterous manipulation skills by learning from human videos, addressing the persistent challenge of noisy trajectory estimation that has hampered prior approaches. While these are early-stage publications with 11 and 2 citations respectively, they position Schmid at the frontier of embodied AI research. Her work reflects a broader mission to make robots more capable, adaptable, and grounded in rich, human-centric visual understanding — a contribution increasingly vital as robotics moves toward real-world deployment.