Junqi Wu

Papers

Total Citations

H-Index

About

Junqi Wu is a researcher at the forefront of 3D perception and adversarial machine learning, with a focus on the security and robustness of point cloud-based systems. His most-cited work, “Gradient-based sparse voxel attacks on point cloud object detection” (2024, 11 citations), introduces a novel method for generating efficient, targeted adversarial perturbations that exploit the sparse voxel structure of LiDAR data. This contribution is critical for understanding vulnerabilities in autonomous driving perception pipelines, where point cloud object detectors must withstand real-world noise and malicious interference. Wu’s approach leverages gradient information to craft sparse, imperceptible attacks that degrade detection performance while minimizing computational overhead, advancing both attack and defense strategies in 3D vision. His research bridges computer graphics, geometric deep learning, and security, offering practical insights for building more resilient autonomous systems. As an emerging voice in the field, Wu’s work has already garnered attention from the computer vision and robotics communities, positioning him as a key contributor to the safe deployment of 3D perception technologies.