Papers
1
Total Citations
17
H-Index
1
About
Zhenbo Shi is a leading researcher in the field of 3D computer vision and adversarial machine learning, with a particular focus on the robustness and security of deep neural networks for 3D point cloud data. His most cited work, "Shape Prior Guided Attack: Sparser Perturbations on 3D Point Clouds" (2022, 17 citations), addresses a critical vulnerability in 3D classification models used in autonomous driving, robotics, and drone navigation. Shi introduced a novel adversarial attack framework that leverages shape priors to generate sparser, more imperceptible perturbations, exposing fundamental weaknesses in current 3D neural network architectures. This research is pivotal for developing more resilient AI systems in safety-critical applications. Beyond this, Shi's broader contributions explore the intersection of geometric deep learning and adversarial robustness, helping to bridge the gap between theoretical security and practical deployment. His work has been recognized for its innovative approach to balancing attack effectiveness with perturbation sparsity, setting a new benchmark for evaluating 3D model integrity. As 3D vision becomes integral to real-world autonomous systems, Shi’s research provides essential insights for building more trustworthy and secure AI.
Research Focus
Key Achievements
Top Papers
- 1Shape Prior Guided Attack: Sparser Perturbations on 3D Point Clouds17 citations · 2022