Kai Leng
Papers
1
Total Citations
3
H-Index
1
About
Kai Leng is a computer vision researcher whose work centers on geometric understanding for robotics and autonomous systems. His primary research areas include relative camera pose estimation, 3D scene understanding, and deep learning for spatial reasoning. Leng’s most notable contribution is the development of SiTPose, a novel siamese convolutional transformer architecture that directly regresses the translation and rotation between two overlapping image frames. This work addresses a fundamental challenge in visual odometry and structure-from-motion, offering an end-to-end learning alternative to traditional geometric pipelines. While his citation count is still building—with his key paper garnering 3 citations to date—the work demonstrates a forward-looking integration of transformer attention mechanisms with convolutional feature extraction for spatial tasks. Leng’s research is particularly relevant for applications in robot navigation, augmented reality, and autonomous driving, where accurate camera pose estimation is critical. His approach of combining siamese network design with transformer architecture represents an innovative step toward more robust and efficient spatial perception systems.
Research Focus
Key Achievements
Top Papers
- 1