Papers
2
Total Citations
21
H-Index
2
About
Zehang Shen is a researcher whose work lies at the intersection of 3D computer vision and autonomous systems, with a primary focus on scene flow estimation—the task of tracking the 3D motion of every point across consecutive point clouds. This capability is foundational for motion perception in autonomous driving and service robotics. Shen’s most notable contribution is the introduction of **SFGAN**, an unsupervised generative adversarial framework that learns 3D scene flow directly from the 3D scene itself, without requiring labeled data. This approach addresses a critical challenge: while sensors like LiDAR and RGB-D cameras capture discrete points, real-world motion is continuous. By leveraging the inherent structure of the scene, SFGAN enables more robust and realistic flow estimation. The work has accumulated over 20 citations, reflecting its relevance in the rapidly advancing field of 3D perception. Shen’s research is particularly valuable for applications where labeled motion data is scarce, offering a path toward scalable, self-supervised learning for dynamic 3D environments.
Research Focus
Key Achievements
Top Papers
- 1
- 2