Yushi Du

Papers

Total Citations

H-Index

About

Yushi Du is a rising researcher in 3D computer vision and geometric deep learning, with a focus on shape assembly and equivariant representations. Their most cited work, "Leveraging SE(3) Equivariance for Learning 3D Geometric Shape Assembly" (2023, 13 citations), introduces a novel framework that exploits SE(3) equivariance—a property ensuring predictions remain consistent under rotations and translations—to reassemble fragmented 3D objects into complete shapes. This contribution addresses a fundamental challenge in geometric part assembly, such as reconstructing broken artifacts or bowl fragments, where traditional methods often struggle with arbitrary orientations. By embedding equivariance into the learning process, Du’s approach improves both accuracy and robustness, enabling more reliable assembly from unordered, unaligned parts. This work has implications for archaeology, robotics, and digital restoration, where automated shape reconstruction is critical. Though early in their career, Du’s research demonstrates a strong commitment to advancing 3D geometric understanding, and their growing citation count reflects the community’s interest in their innovative methods. Their work stands out for bridging theoretical equivariance principles with practical assembly tasks, marking them as a promising contributor to the field.