Zhengqi Li
Papers
1
Total Citations
6
H-Index
1
About
Zhengqi Li is a leading researcher in computer vision, specializing in 3D scene understanding, dynamic motion estimation, and novel view synthesis. His work bridges the gap between static 3D reconstruction and the complex, moving world, with a focus on learning spatial-temporal representations from real-world imagery. Li’s major contributions include pioneering methods for recovering 3D motion and structure from video, most notably through his work on "Stereo4D" (2025, 6 citations), which introduces a framework for learning how objects move in 3D directly from internet stereo videos—a task that has traditionally resisted large-scale supervised training due to the difficulty of obtaining ground-truth motion labels. This approach enables more robust dynamic scene reconstruction, with implications for robotics, autonomous navigation, and augmented reality. Li’s research has garnered attention for its innovative use of self-supervised learning and multi-view geometry, pushing the boundaries of what can be inferred from unlabeled video data. His work is foundational for advancing 4D scene understanding, making him a key figure in the next generation of spatial AI research.
Research Focus
Key Achievements
Top Papers
- 1Stereo4D: Learning How Things Move in 3D from Internet Stereo Videos6 citations · 2025