Yan Di

Papers

Total Citations

H-Index

About

Yan Di is a rising star in computer vision and robotics, whose research focuses on bridging the gap between 3D perception and embodied AI. His core contributions lie in 6-DoF robotic grasping, object rearrangement, and category-level object pose estimation. Di’s work on MonoGraspNet (2023, 42 citations) tackles the long-standing challenge of 6-DoF grasping from a single RGB image, moving beyond depth-dependent methods to handle photometrically challenging objects with superior accuracy. In SG-Bot (2024, 24 citations), he introduced a coarse-to-fine robotic imagination framework using scene graphs for object rearrangement, a key capability in embodied AI that enables robots to understand and manipulate complex environments. Most recently, his work on SE(3)-equivariance learning (2025) advances category-level object pose estimation by addressing the limitations of direct regression from point clouds, offering more robust vision-based measurement for robotics. Di’s research is notable for its practical impact on real-world robotic manipulation, with his papers already garnering significant attention in the community. His innovative use of geometric deep learning and scene-level reasoning positions him as a leading voice in the next generation of intelligent robotic systems.