Zhenbo Xu

University of Science and Technology of China, Beihang University

Papers

Total Citations

102

H-Index

About

Zhenbo Xu is a leading researcher in 3D computer vision, with a primary focus on 3D object detection for autonomous driving and robotics. His most impactful contribution is the development of **ZoomNet**, a part-aware adaptive zooming neural network that addresses the critical challenge of accurately estimating the 3D pose of distant and occluded objects from stereo imagery. This work, which has garnered over 85 citations, significantly advances the reliability of perception systems in real-world, cluttered environments. Beyond detection, Xu has made important strides in the security of 3D deep learning models. His work on **Shape Prior Guided Attacks** introduces a novel method for generating sparse, efficient adversarial perturbations on 3D point clouds, exposing vulnerabilities in classification models used across robotics and autonomous systems. By combining high-performance detection with rigorous robustness analysis, Xu’s research not only pushes the boundaries of what 3D vision systems can perceive but also ensures they are resilient against malicious inputs. His contributions are essential reading for anyone working at the intersection of 3D perception, safety, and adversarial machine learning.