Gedas Bertasius

California University of Pennsylvania, University of North Carolina at Chapel Hill

Papers

Total Citations

H-Index

About

Gedas Bertasius is a computer vision and robotics researcher whose work spans egocentric perception, imitation learning, and scalable robot learning. He is perhaps best known for his pioneering contributions to first-person action-object detection, most notably through EgoNet, which investigates the nuanced interplay between human visual attention and motor interactions with objects as captured from a first-person camera perspective. This foundational work, which has accumulated over 44 citations, challenged conventional third-person robotic vision paradigms by leveraging the intimacy and richness of egocentric viewpoints to better understand human sensorimotor behavior. More recently, Bertasius has extended his research toward practical challenges in robot learning, exploring how scalable demonstration collection can be achieved through augmented reality pipelines, as demonstrated in his ARCADE framework. His 2025 work on ReBot further reflects his commitment to bridging the gap between simulation and real-world robotics by synthesizing robotic video data to enhance vision-language-action model training without prohibitive data collection costs. Collectively, his research reflects a coherent trajectory from understanding human perception to empowering robots with scalable, human-inspired learning, making his work highly relevant for researchers working at the intersection of computer vision, embodied AI, and robot learning.