Kexin Zhang

Nanjing University of Posts and Telecommunications

Papers

Total Citations

H-Index

About

Kexin Zhang is a rising researcher in 3D computer vision, with a focus on point cloud perception for autonomous driving and robotics. Their work tackles the critical challenge of balancing accuracy and speed in 3D object detection and tracking. Zhang’s most cited paper, “HCPVF: Hierarchical Cascaded Point-Voxel Fusion for 3D Object Detection” (2023, 30 citations), introduces an innovative architecture that fuses point-based and voxel-based representations in a cascaded manner, achieving state-of-the-art detection performance while maintaining real-time efficiency. This work addresses a fundamental trade-off in the field and has quickly gained recognition. In addition, Zhang contributed “Accurate 3D Single Object Tracker in Point Clouds with Transformer” (2022), which adapts the successful transformer architecture from 2D tracking to 3D point clouds, pioneering a new direction for single object tracking. By developing Trans3DT, Zhang demonstrated how attention mechanisms can effectively model spatial relationships in sparse 3D data. Though early in their career, Zhang’s work on hierarchical fusion and transformer-based tracking is shaping the next generation of efficient, accurate 3D perception systems for autonomous vehicles and intelligent robots.