Haozhe Du

Papers

Total Citations

H-Index

About

Haozhe Du is a rising researcher in computer vision and robotics, whose work centers on advancing pose registration—a fundamental challenge for enabling machines to perceive and interact with 3D environments. His most notable contribution, "DPCN++: Differentiable Phase Correlation Network for Versatile Pose Registration" (2023), tackles the difficult problem of initialization-free pose registration up to 7 degrees of freedom, applicable to both homogeneous and heterogeneous measurements. This work introduces a novel differentiable solver that overcomes the limitations of prior learning-based methods, which often depend on heuristic initialization or restrictive assumptions. By leveraging phase correlation techniques within a neural network framework, Du’s approach achieves robust performance without requiring pre-aligned inputs, marking a significant step toward more flexible and reliable registration systems. Though early in his career, with 4 citations on this key paper, his research addresses a critical bottleneck in vision and robotics, promising to enhance applications from autonomous navigation to 3D reconstruction. Du’s innovative integration of differentiable optimization with geometric reasoning positions him as a promising contributor to the next generation of perception systems.